An intersample peak is a hidden analogue peak that forms between two digital samples during playback reconstruction, and it can exceed 0 dBFS even when every sample in your file reads safely below it. That hidden peak can cause distortion at the digital-to-analogue stage or after your file goes through lossy encoding. The fix is simple: measure true peak, not just sample peak, and master to a conservative ceiling of −1 to −2 dBTP.
TL;DR:
- EBU R128 recommends a −1 dBTP production ceiling, while −2 dBTP provides extra headroom when lossy encoders may raise peaks.
- Dense electronic music, metal, and heavily limited masters face greater risk because sharp transients and high frequency energy encourage reconstruction overshoots.
- After limiting, recheck true peak following sample rate conversion or dithering, and test a short excerpt through the target codec and bitrate before delivery.
Table of Contents
- How intersample peaks are created in digital audio
- How true peak measurement actually works
- Why intersample peaks cause real problems after delivery
- Step‑by‑step: preventing intersample peaks before export
- What to look for in meters and limiters
- How AubioMix helps you spot true‑peak issues before they ship
- A modest ceiling buys you a lot of reliability
- Catch true‑peak risks automatically with AubioMix
- FAQ
- Sources
How intersample peaks are created in digital audio
To understand where these peaks come from, we need to go back to how digital audio actually works. Under the Nyquist‑Shannon sampling theorem, your DAW stores a series of discrete sample values, snapshots of the waveform at fixed intervals, rather than the continuous curve you hear. When you hit play, your converter has to rebuild that continuous curve from those dots, using a reconstruction filter that interpolates between them.
Here's the catch: the smooth curve a reconstruction filter draws between two samples doesn't have to stay within the amplitude of those samples. It can swing higher, sometimes well above 0 dBFS, even though neither sample value was clipping. Your meter never sees this because it only reads the stored sample values, not the rebuilt waveform.
Some material is far more prone to this than others. Transients, snare hits, kick attacks, anything with a sharp, fast‑rising edge, tend to produce steeper curves between samples, which gives the reconstruction filter more room to overshoot. High‑frequency content behaves similarly, since it sits closer to the Nyquist limit where the gap between what's sampled and what's reconstructed is largest. A heavily limited master full of hard transients is prime territory for intersample peaks, which is exactly why mastering engineers need to measure for them specifically, not just assume a clean sample peak reading tells the whole story.
How true peak measurement actually works
Sample-peak metering checks only the stored samples' amplitude and misses overshoots between them. True-peak metering oversamples the signal and interpolates between samples to approximate the reconstructed analogue waveform, revealing hidden peaks.
The industry standard, ITU‑R BS.1770, specifies oversampling at a minimum of four times the original sample rate for reliable true‑peak detection, and the standard notes that higher oversampling ratios improve measurement accuracy further. The filter doesn't need full audio‑quality reconstruction either. BS.1770 allows shorter FIR filters and reduced word‑length in the oversampling stage, since the goal is an accurate peak reading rather than a listenable signal.
EBU R128 builds on this by tying true‑peak measurement into loudness workflows directly, recommending a maximum production true‑peak level of −1 dBTP and offering its own implementation notes on headroom and oversampling. Between the two standards, you've got a clear, tested framework for what a compliant meter should be doing under the hood, which matters when you're choosing which tools to trust.
Why intersample peaks cause real problems after delivery
The danger with intersample peaks is that they're invisible until the moment they're not. Your DAC reconstructs the continuous waveform every time your file plays back on a real device, so a peak that looked safe on a sample‑peak meter can still clip at that reconstruction stage, producing audible distortion you never heard in the studio.
Lossy encoding makes this worse. Converting to MP3 or AAC reshapes the waveform during compression, and that process can push true‑peak levels higher than they were in your original file, sometimes creating clipping where none existed before encoding according to Tonalux's technical breakdown.
Dense electronic music, metal and anything mastered with heavy limiting are particularly exposed, since these styles rely on hard transients and high‑frequency energy pushed close to the ceiling. Most streaming platforms expect files delivered with adequate headroom, so true-peak checks before upload help avoid distortion caused by encoding.
Step‑by‑step: preventing intersample peaks before export
Catching these peaks doesn't require guesswork, just a consistent check at each stage of your export chain.
- Measure with a true‑peak compliant meter. Confirm it oversamples at a minimum of 4x, per BS.1770 guidance, before you trust any reading.
- Insert a true‑peak aware limiter with lookahead. Let it process in the oversampled domain, then re‑measure your output.
- Set your delivery ceiling. Use −1 dBTP as a baseline, or drop to −2 dBTP for extra safety margin across tighter encoders, and recheck true peak after any sample‑rate conversion or dithering step.
- Tame remaining transients. If peaks persist, pull back a touch of limiting or use a transient shaper to soften the sharpest edges, then run the measurement again.
Pro Tip: Export a two‑ or three‑second snippet around your loudest moment, run it through the actual codec and bitrate your target platform uses, then re‑measure true peak on that encoded file, since this catches the headroom‑eating effect of lossy encoding before it reaches listeners.
What to look for in meters and limiters
Not every plugin labelled "true peak" is built to the same standard, so it's worth checking the specifics before you rely on one for final delivery.
- Declared oversampling ratio: look for 4x as a floor, with some tools offering 8x or 16x for tighter accuracy.
- BS.1770 compliance: stated explicitly in the plugin documentation, not just implied by a "true peak" label.
- dBTP readout with gating options: lets you isolate the true‑peak value from an integrated loudness reading during the same pass.
- Limiter lookahead and oversampled processing: a true‑peak limiter needs both to catch and tame peaks before they're printed.
- Transparent behaviour on attack and release: oversampled processing should not introduce audible artefacts of its own.
One deviation worth knowing: ITU‑R BS.1770 notes that sample‑peak meters can under‑read true peaks by several decibels on high‑frequency transient material, which is the gap a proper oversampling meter is built to close.
How AubioMix helps you spot true‑peak issues before they ship
Checking all of this by ear or by eye, across every export, gets tedious fast. Our platform runs automated true‑peak and loudness checks on any mix you upload, comparing the results against genre benchmarks so a dense techno master isn't judged by the same ceiling as a ballad. When a report flags a true‑peak violation, it comes with a concrete next step, whether that's easing off the limiter or attenuating a specific transient, alongside our audio metrics glossary for anyone who wants the deeper technical grounding.
A modest ceiling buys you a lot of reliability
Chasing the last half a decibel of loudness is rarely worth the risk of distortion on a listener's phone speaker or car stereo. We'd rather see a mix sit at a slightly lower true peak and translate cleanly everywhere than win a loudness war and lose on playback. Make true‑peak verification a routine part of final QC, not an afterthought.
— AubioMix
Catch true‑peak risks automatically with AubioMix
Running true‑peak and loudness checks by hand on every mix, then rechecking after encoding, eats time most of us would rather spend mixing. Upload your file to AubioMix and our report flags true‑peak violations against genre benchmarks, with specific steps for fixing them, whether that means easing a limiter setting or softening a transient, before you commit to a master.
The workflow is straightforward: upload your track, get your report, apply the suggested fixes, then re‑upload to confirm the issue is cleared. A single Mix Review costs £4.99, or you can run checks regularly on a Professional plan at £19.99 a month. Check our pricing page to find the option that fits how often you master.
FAQ
What is the difference between sample peak and true peak?
Sample peak only measures the amplitude of the stored digital samples themselves. True peak oversamples the signal and reconstructs the likely analogue waveform between those samples, which is how it reveals intersample peaks that a sample‑peak meter misses entirely.
What should my true peak be when mastering?
EBU R128 recommends a maximum production true‑peak level of −1 dBTP. Many mastering engineers drop to −2 dBTP for extra safety margin when a file will pass through lossy encoding before reaching listeners.
What does audio peak mean?
An audio peak is the highest amplitude point in a signal at a given moment. In digital audio this can mean the highest stored sample value, or, more usefully, the highest point the reconstructed analogue waveform reaches during playback.
Where should my master peak?
A true‑peak ceiling of −1 dBTP works for most delivery targets, with −2 dBTP giving extra headroom for platforms using tighter encoders. Always recheck your true peak after any sample‑rate conversion or dithering, since those steps can shift the reading.
Sources
For implementation detail beyond this guide, ITU‑R BS.1770 covers the oversampling algorithm itself, while EBU R128 ties true‑peak limits into loudness workflows. For a worked explanation of reconstruction and limiting behaviour, see our true‑peak mastering guide.
- Recommendation ITU‑R BS.1770‑5 (11/2023) Algorithms to measure audio programme loudness and true‑peak audio level
- Loudness normalisation and permitted maximum level of audio signals (EBU R128)
- The Phantom Peak: Why samples below 0 dBFS still cause distortion | Tonalux blog