Unveil

Free online true peak meter

Upload a master and get its true peak in dBTP, measured with 4× polyphase oversampling per ITU-R BS.1770-4 Annex 2 — the inter-sample peaks that a normal sample peak meter misses and that clip audibly after MP3/AAC encoding. Aim for −1.0 to −0.1 dBTP for a clean streaming master.

Sample peak versus true peak

A digital file stores a sequence of samples. The sample peak is simply the largest sample value, and that is what a basic peak meter or a waveform display shows. But the sound you hear comes from the reconstructed continuous waveform that a digital-to-analogue converter draws between those samples. That curve can rise above the highest sample. The highest point of the reconstructed waveform is the true peak, and it is reported in dBTP (decibels true peak) relative to full scale.

The difference matters because a file can sit at -0.1 dBFS on the sample meter and still contain a reconstructed peak above 0 dBFS. When that happens, the converter or a downstream codec has no headroom for it, and clipping occurs even though your meter never showed red.

Inter-sample peaks: how large can the error be

An inter-sample peak occurs when two adjacent samples straddle the true crest of the waveform. A clean demonstration: take a sine wave at one quarter of the sample rate, with a phase offset of 45 degrees. Every sample then lands at about 0.707 of the amplitude, so the sample peak reads about -3 dBFS. The underlying sine, however, reaches full amplitude, so the true peak is 0 dBTP. That is a 3 dB error from a perfectly ordinary signal.

Real music rarely hits that exact worst case, but heavily limited masters are the likeliest to show overs of 0.5 to 2 dB or more, because their waveforms are flattened against the ceiling and many consecutive samples sit near full scale. The more aggressively a master is pushed to 0 dBFS, the more likely it is to hide inter-sample overs.

How true peak is measured: oversampling per BS.1770

ITU-R BS.1770-4 Annex 2 defines a true-peak meter. The signal is upsampled by an integer factor with an interpolation filter, which gives you the values between the original samples, and the highest absolute value in the upsampled stream is taken as the true peak. For material at 44.1 or 48 kHz, the standard describes 4x oversampling; that is what most meters use, and it captures nearly all of the overshoot. Higher rates such as 96 kHz need less oversampling because the samples are already closer together.

Oversampling is an estimate, not an absolute truth: a different reconstruction filter in a different converter can produce slightly different peaks. This is one reason a safety margin is used rather than a ceiling of exactly 0 dBTP.

Codec overshoot after MP3, AAC and Ogg

Lossy codecs work by discarding information the ear is least likely to notice and quantizing the rest. This does not preserve the waveform's peak relationship. After decoding, the signal usually has slightly different peak levels than the source, and the change is not always downward. A master with a 0 dBTP ceiling can decode to peaks above full scale, which then clip in the playback chain.

How much overshoot you get depends on the codec, the bitrate, and the material. Lower bitrates and dense, loud material tend to produce more. Treat figures like a fraction of a dB to more than a dB as a plausible range rather than a fixed constant, and verify by encoding your own master and measuring the decoded result.

This is why the common recommendation is to leave margin below 0. Streaming services transcode your file, often to more than one format, and you do not control the decoder.

Choosing a ceiling and setting the limiter

A ceiling of about -1 dBTP is the conventional safe choice. It is also the maximum in EBU R128 for broadcast. Some distributors suggest lower values, around -2 dBTP, for very loud masters because the louder the source, the more codec overshoot it tends to produce. Check the guidance from the platform you deliver to, since it can change.

Limiter settings that affect true peak behaviour:

Step-by-step: checking and fixing true peak

  1. Bounce the final master at its delivery sample rate and bit depth.
  2. Run a true-peak meter on the whole file, not a section. Note the maximum dBTP.
  3. Compare to your ceiling. Anything above -1 dBTP needs a fix; anything above 0 dBTP will clip on conversion.
  4. Locate where the highest peaks occur. They are usually kicks, snares, or the loudest chorus downbeat.
  5. Lower the limiter ceiling or enable true-peak mode and re-bounce. Do not simply lower the output fader afterwards, as that changes loudness along with peak.
  6. If you need to keep loudness, reduce the offending transient earlier in the chain with a clipper or a short-attack compressor, then let the limiter do less work.
  7. Encode a test MP3 or AAC, decode it, and measure again. If it still goes above 0 dBFS, lower the ceiling by another 0.5 dB.

Worked example and checklist

A master reads -0.2 dBFS on the sample peak meter but +0.9 dBTP on a true-peak meter. That is a 1.1 dB inter-sample excess. A 4x oversampling meter has found a reconstructed crest above full scale. Setting the limiter to a -1.0 dBTP ceiling with true-peak detection brings the maximum down to -1.0 dBTP, a reduction of 1.9 dB at the peak. Because the integrated loudness depends mostly on the body of the signal and the limiter only touches the highest crests, integrated loudness might fall by only 0.3 to 0.5 LU. A test AAC encode then decodes to about -0.6 dBFS, within the safe zone.

What you get

Questions

What true peak should a master hit?

−1.0 to −0.1 dBTP is the clean target — enough headroom that lossy encoding won't push inter-sample peaks into clipping.

Why does true peak differ from sample peak?

Sample peak only reads the existing samples; true peak reconstructs the analog waveform between them, where the real maximum often sits higher.

Related

Analyze your mix — free