True peak vs sample peak: a practical limiting guide

Sample peak measures the highest stored sample value. True peak estimates the highest level of the waveform reconstructed between those samples. The reconstructed curve can exceed the samples, so a master that never reaches 0 dBFS on a sample meter can still produce a higher peak during playback, sample-rate conversion or encoding.

That does not mean every inter-sample peak will create audible distortion. It means the sample meter alone cannot describe the entire reconstruction.

Why peaks can exist between samples

Digital audio stores measurements at discrete moments. A digital-to-analogue converter does not play those measurements as a staircase; it reconstructs a continuous signal from them.

The reconstructed curve can bend above two neighbouring sample values. ITU-R BS.1770-5 explicitly notes that actual signal peaks usually occur between sampling instants and recommends oversampling for true-peak estimation.

Think of four points drawn on a smooth hill. None of the points must sit on the summit for the curve connecting them to pass above every point.

What a true-peak meter does

A true-peak meter oversamples the signal through a reconstruction filter and looks for peaks in the denser result. The reading is expressed in dBTP: decibels relative to full scale under true-peak measurement.

It is still an estimate. Filter design and oversampling factor affect accuracy. A limiter may use one detector internally while a separate delivery meter uses another. Leave sensible headroom instead of treating the final decimal as an absolute guarantee.

Sample clipping, inter-sample overs and audible distortion

These ideas are related but not identical:

• Sample clipping occurs when stored sample values hit or exceed the representable limit.
• An inter-sample over is a reconstructed peak above full scale even though the samples themselves may remain below it.
• Audible distortion depends on what the converter, codec or later process does with that signal.

A true-peak warning identifies risk. It does not prove that every listener will hear clipping.

Why encoding can change the peak

Lossy encoding changes the waveform. The decoded result can have different sample and true-peak values from the source master. Sample-rate conversion and some other processing can also expose peaks that the original sample meter did not show.

This is why delivery headroom matters even when the exported lossless file appears technically below 0 dBFS.

Choosing a limiter ceiling

Follow the specification for the actual destination. There is no single ceiling for all music, broadcast, video and archive masters.

Spotify currently recommends keeping masters below -1 dBTP, and recommends below -2 dBTP when the master is louder than -14 LUFS, because louder material can be more susceptible to distortion during encoding. EBU R 128 is a broadcast recommendation with a different context and target; it should not be presented as a universal music-streaming prescription.

A practical music workflow is:

1. Find the delivery requirements.
2. Enable a reliable true-peak meter.
3. Set the limiter ceiling with enough margin for that destination.
4. Export at the intended sample rate and bit depth.
5. Re-import and measure the actual export.
6. Audition a delivery-codec preview when available.

When true-peak limiting is useful

Use true-peak detection near the end of a delivery chain when:

• a platform or broadcaster specifies a dBTP limit;
• the file will be lossily encoded;
• the master is dense and close to full scale;
• sample-rate conversion is part of delivery;
• you need a defensible maximum-level measurement.

It is less urgent on an individual mix channel with substantial headroom. Enabling high oversampling everywhere can also increase CPU use or latency.

Mistakes to avoid

• Setting the ceiling to 0 dBTP and assuming conversion cannot change anything.
• Confusing true-peak compliance with good sound.
• Driving a limiter harder merely because true-peak mode is enabled.
• Comparing a louder limited version against a quieter original.
• Ignoring the exported file and measuring only the live session.
• Applying a streaming recommendation as if every destination shared it.

A clean mastering check

Level-match the limited and unprocessed versions. Listen for softened transients, pumping, low-end modulation and high-frequency grit. Then inspect integrated loudness, loudness range, gain reduction and true peak together.

SPC Limit Breaker provides true-peak lookahead limiting with 4× inter-sample detection, eight limiting characters, loudness metering, and Delta and Unity monitoring. Use SPC SpectraScope after the chain for independent BS.1770-5 loudness and true-peak inspection without processing the signal.

Next, learn why a ceiling alone cannot answer how loud your master should be.