True peak vs sample peak: why −1 dBTP, and when −2 is safer

The short answer. A sample-peak meter only sees your samples. The waveform your DAC and the streaming encoder rebuild can go higher between them. True peak estimates those in-between peaks. For streaming, keep true peak below −1 dBTP, and below −2 dBTP if your master is louder than −14 LUFS. Use a true-peak limiter to get there.

What a sample-peak meter misses

A digital file is a list of samples. When it's played, a converter draws a smooth curve through them, and that curve doesn't have to turn around exactly on a sample. If two samples sit either side of a crest, the real peak is between them, and higher than both.

The standard for loudness measurement, ITU-R BS.1770, defines it this way: "True-peak level is the maximum (positive or negative) value of the signal waveform in the continuous time domain; this value may be higher than the largest sample value in the 48 kHz time-sampled domain."

It also says how far off a plain peak meter can be:

• for steady tones, "a 3 dB under-read for an unfortunately-phased tone at a quarter of the sampling frequency";
• for real music, "the higher the frequency content of the transient, the larger the potential under-read", and "under-reading of these can commonly be several dBs".

That's why a snare crack or a bright clap can read −0.3 dBFS on a sample meter and still clip a converter.

How true peak is measured

BS.1770's annex describes the method: oversample the signal (4x, from 48 kHz to 192 kHz), low-pass filter it, take the absolute value, and convert to dB. The result is in dBTP, decibels relative to full scale, true-peak.

Oversampling is not guessing. As the annex puts it, sampling theory shows the in-between waveform "can be done, because it is known that the sampled signal contains no frequencies above half of the original sampling frequency".

A true-peak meter is still an estimate. BS.1770's own table puts the worst case for 4x oversampling at an under-read of 0.688 dB, for content right at the top of the band. The AES streaming guidance (TD1008) says true-peak meters "typically have an error of less than 0.6 dB". That alone is a reason not to master to 0.0 dBTP.

Why peaks go up after you export

Your WAV is not what listeners hear. Several steps on the way can push peaks higher:

So the ceiling isn't about your file. It's about the room your file needs after everything downstream has had a go at it.

What the platforms and standards say

Spotify's extra warning about loud masters, from Track not as loud as others?: "Very loud masters with true peaks above -2 dB can introduce distortion during encoding, which can increase perceived loudness — but not necessarily improve quality."

−1 or −2?

• −1 dBTP is the default for streaming at around −14 LUFS. Every source above agrees on it.
• −2 dBTP if your master is louder than −14 LUFS. Loud, dense masters overshoot more in the encoder, and Spotify says so explicitly.
• −0.5 dBTP is the ceiling monody's Club target uses, for masters DJs play from lossless files. AES notes high-bit-rate coders can cope with it, but for streaming it's cutting it fine. See Streaming master vs club master.

Remember that on Spotify the extra decibel of loudness you'd gain by pushing the ceiling up is turned back down by normalization anyway. More in How loud should your master be? and Does Spotify turn quiet songs up?.

Setting a true-peak limiter

1. Put it last. The AES guidance: "always include a true-peak limiter as the last element before the codec". Nothing after it should add gain or filter.
2. Turn on true-peak (or oversampling) mode if the limiter has one. Without it, the limiter holds samples, not the waveform between them.
3. Set the ceiling below the target. −1.0 dBTP for streaming, −2.0 for loud masters.
4. Measure the result with a true-peak meter, over the whole song. A limiter's ceiling is a target, not a guarantee, especially with a fast release.
5. Watch the PLR (true peak minus integrated loudness). A common rule of thumb: below about 6 dB, most music starts to sound squashed.

How monody does it

monody's Finish view measures the rendered song's true peak with 4x oversampling and shows it against the target's ceiling: Streaming −1 dBTP, Apple Music −1 dBTP, Club −0.5 dBTP. For another ceiling, ask the assistant: it can measure and master to a custom target. It also shows integrated loudness, short-term maximum, loudness range and PLR. A song that peaks over the ceiling gets a Fix finding.

Master to target puts monody's Maximizer on the master bus with its true-peak option on, sets the ceiling just under the target's, solves the drive on a render, then measures a full render and corrects if the loudness or a peak is still off. It's loudness and peak control, as one undoable step, with no EQ.

Sources

• ITU-R: BS.1770-5, Algorithms to measure audio programme loudness and true-peak audio level (Annex 2 and its Attachment)
• EBU: R 128 (PDF)
• AES: TD1008, Recommendations for loudness of internet audio streaming and on-demand distribution (PDF)
• Spotify for Artists: Loudness normalization and Track not as loud as others?
• Apple: Apple Digital Masters (PDF)

Checked on 1 October 2026.