/piːk/
Peak is the highest instantaneous amplitude reached by an audio signal at any given moment. It is measured in dBFS on digital meters and determines whether a signal clips or retains clean headroom for further processing.
Every distorted master, every clipped vocal that never got fixed in the mix — it all started with a meter that nobody watched closely enough.
In audio production, peak refers to the highest instantaneous amplitude that an audio signal reaches at any discrete moment in time. Unlike loudness measurements such as RMS or LUFS — which average energy across a window of time — a peak value captures the absolute maximum excursion of a waveform sample by sample, with no smoothing or integration applied. In the digital domain this value is expressed in decibels relative to full scale (dBFS), where 0 dBFS represents the absolute ceiling of a digital system; any signal that reaches or exceeds this threshold produces hard clipping, an abrupt, mathematically certain form of distortion caused by the inability of the system to represent values beyond its numerical range.
Understanding peak level is foundational to every decision a producer makes about gain structure. When you set an input gain on an audio interface, adjust a channel fader in a DAW, or choose a threshold on a limiter, you are ultimately negotiating with peak values. The difference between the loudest peak in a signal and the system's clipping point is called headroom — and headroom is the single most important resource you manage throughout a session. Leaving adequate headroom at the recording stage means that unexpected transients — the sharp attack of a snare rim shot, a vocalist's sudden consonant burst, a guitarist's unexpected hammer-on — have room to exist without distorting. Allowing peaks to reach 0 dBFS in a recording session forfeits the ability to process that signal cleanly afterward, because many dynamics processors, EQ algorithms, and saturation plugins generate inter-stage gain that pushes already-maxed signals further into clipping.
The relationship between peak and perceived loudness is frequently misunderstood by producers entering the field. A signal can carry enormous peak values while sounding quiet if those peaks are sparse — a single isolated snare hit in an otherwise empty arrangement, for instance. Conversely, a dense mix of mid-range instruments with limited dynamic range may read relatively low on a peak meter while measuring high on an RMS or integrated LUFS meter. This distinction is not academic: streaming platforms such as Spotify, Apple Music, and YouTube normalize playback according to integrated loudness targets (typically −14 LUFS or −16 LUFS), and they do so without regard for peak levels. A master with crushing peak-limited dynamics may be turned down to meet the platform target, and the listener receives it quieter — not louder — than a master with natural dynamic range. Peak metering tells you where your ceiling is; LUFS metering tells you how loud the music actually sounds. Both are necessary, and neither replaces the other.
Analog audio systems handle peak excursions differently than digital ones. Magnetic tape and analog console circuitry exhibit graceful, musically coherent saturation as signals exceed nominal operating levels — a behavior that has been the subject of enormous reverence in recording culture and the basis for an entire category of plugin emulation. Digital systems, by contrast, clip with zero tolerance and zero musicality the instant a sample value reaches or exceeds the maximum representable integer. This binary nature of digital clipping is why peak awareness is more critical in digital production than it was in tape-era work: there is no soft shoulder, no gentle compression into saturation. There is signal, and then there is clipping, and the boundary between them is exact and unforgiving. This is also why the concept of true peak matters: because digital-to-analog conversion involves reconstruction filtering that can generate inter-sample peaks above 0 dBFS even when no individual sample in the digital domain reaches that ceiling. True peak measurement accounts for this reconstruction, making it the standard required for broadcast and streaming delivery specifications.
At the sample level, a digital audio signal is a sequence of discrete numerical values — in a 24-bit system, integers ranging from −8,388,608 to +8,388,607. Each value represents the instantaneous displacement of the virtual membrane, the voltage at an interface's input, or the computed output of a plugin's algorithm at a single moment in time. A peak detector, whether in hardware or software, watches this stream continuously and records the maximum absolute value it encounters. In a simple peak meter, this captured maximum is displayed — often with a hold function that freezes the highest value on screen for several seconds — so the engineer can evaluate whether the signal is approaching the clipping ceiling. The measurement is expressed in dBFS: 0 dBFS equals the maximum representable value; −6 dBFS is half that amplitude on a linear scale (not half the perceived loudness, because the decibel scale is logarithmic); −18 dBFS is roughly the nominal operating level recommended for recording digital audio.
Sample-peak metering, while standard in virtually every DAW, has a known limitation: it only measures values at the discrete sampling instants. When a digital audio file is converted back to analog by a DAC, the reconstruction filter generates a continuous waveform by interpolating between those samples. This interpolated waveform can produce peaks between sample points — so-called intersample peaks or inter-sample overs — that exceed 0 dBFS in the analog reconstruction even though no individual sample in the digital file reached that level. Streaming services and broadcast infrastructure re-encode audio through codecs (AAC, MP3, Opus) that alter sample values during compression, and these re-encoded versions can produce intersample peaks substantially higher than the original. A master delivered at a sample peak of −0.1 dBFS may produce intersample peaks of +1.5 dBTP (decibels relative to true peak) or higher after lossy encoding. This is why broadcast standards such as EBU R128 and streaming platform specifications from Apple Music and Spotify mandate a true peak ceiling of −1 dBTP, and why mastering engineers use true peak limiters rather than simple peak limiters for final delivery.
In a mixing context, peak level interacts with every dynamics processor in a signal chain. A compressor's threshold is compared against the signal's level continuously; when peak values rise above the threshold, gain reduction is applied at a rate determined by the ratio. The attack parameter governs how quickly the compressor responds once a peak crosses the threshold — a slow attack allows transient peaks to pass through unaffected (often desirable for punch), while a fast attack clamps down on peaks immediately (useful for taming harsh consonants on vocals). A limiter is simply a compressor with an extremely high ratio — typically 10:1 or infinity:1 — with a fast attack designed specifically to prevent peaks from exceeding the ceiling. Clippers, which have become a prominent mastering tool in recent years, differ from limiters in that they truncate sample values that exceed a ceiling rather than applying gain reduction before the fact, generating odd-order harmonic distortion products that many engineers find musically useful in limited doses.
The metering ballistics of peak detectors are also worth understanding technically. True peak meters implement an oversampling process — typically 4× or 8× the native sample rate — to reconstruct the inter-sample waveform and measure its actual maxima. This oversampling adds a small amount of computational overhead but is essential for accurate delivery metering. Some meters offer a "peak hold" function that maintains the highest registered value on screen, sometimes with a user-configurable decay time, allowing engineers to check the absolute maximum reached during a full playthrough without watching the meter continuously. Hardware meters in classic consoles used VU ballistics with a 300-millisecond integration time — far too slow to catch sharp transients — which is why VU meters read subjectively lower than peak meters on the same signal, and why mixing "to VU" on a console requires understanding that true peaks are running considerably higher than what the needle shows.
The practical workflow implication of all this physics is straightforward: record with peaks landing between −18 dBFS and −10 dBFS on individual tracks, leave room on the master bus for the mix to breathe and for mastering processors to work, and use true peak limiting at the final mastering stage to ensure delivery compliance. Peaks are not enemies to be suppressed — they are the dynamic information that makes music feel alive. The producer's craft lies in deciding which peaks to preserve, which to tame, and how to arrange all of them so the ceiling of the delivery format never becomes the ceiling of the music.
Diagram — Peak: Diagram showing sample peak versus true peak (intersample peak) on a digital waveform, with headroom, 0 dBFS ceiling, and nominal level annotated.
Every peak — hardware or plugin — operates on the same core parameters. Know these and you can work with any implementation.
Set in dBFS (sample peak) or dBTP (true peak), the ceiling is the hard upper boundary enforced by a limiter or clipper. For streaming and broadcast delivery, −1 dBTP is the standard recommended ceiling, ensuring that lossy codec re-encoding does not generate intersample overs above 0 dBFS. In mixing, individual channels are typically kept below −6 dBFS to preserve bus headroom.
Most DAW peak meters display a hold segment that remains at the maximum detected level for a user-configurable duration — commonly 1 to 3 seconds — before decaying back toward the current signal level. Longer hold times (3–5 s) are useful during a full mix playback review; shorter times (0.5–1 s) are preferred for real-time tracking where the engineer needs immediate feedback. Some meters offer infinite hold until manually reset.
True peak meters oversample the incoming signal — typically 4×, 8×, or 16× the native sample rate — to reconstruct the inter-sample waveform and measure its actual maxima. A 4× oversampled true peak meter on a 44.1 kHz session processes at an equivalent 176.4 kHz internally. Higher oversampling factors yield more accurate intersample peak values but increase CPU load; 4× is considered sufficient for most delivery purposes, while 8× is standard in broadcast-grade metering tools.
Peak mode responds instantaneously to every sample value, making it ideal for catching transients and verifying headroom. RMS mode integrates energy over a window (typically 300 ms to 600 ms), giving a reading closer to perceived loudness. VU ballistics mimic the 300 ms integration of a physical VU meter needle, historically calibrated so that 0 VU equals +4 dBu in professional analog systems. On a digital channel peaking at −10 dBFS, a VU meter may read −18 VU on the same signal — the gap between peak and VU reads indicates dynamic range.
The inter-sample peak margin is the gap between the delivered sample peak ceiling and 0 dBFS, deliberately chosen to absorb the overshoot that occurs during DAC reconstruction and lossy codec processing. Empirical testing by the Audio Engineering Society (AES) and streaming platform engineers has shown that −1 dBTP provides sufficient margin for AAC and MP3 at standard bitrates. Mastering for CD delivery (16-bit, no lossy encoding) can tolerate a tighter margin of −0.5 dBTP, though many engineers use −1 dBTP universally for simplicity.
Every professional DAW and hardware meter includes a clip indicator — typically a red segment or LED — that activates whenever a sample value reaches 0 dBFS (in some implementations, when two or more consecutive samples reach 0 dBFS). The indicator is latching by design so that a single clipped sample that occurred during a long take does not go unnoticed. Checking the clip indicator after every record pass is a non-negotiable habit; a latching clip on an input channel during tracking means the recorded file is permanently damaged at that moment.
Session-ready starting points. Values are starting-point targets; always verify with a calibrated true peak meter before delivery, and adjust for genre dynamic range requirements.
| Parameter | General | Drums | Vocals | Bass / Keys | Bus / Master |
|---|---|---|---|---|---|
| Recording peak target | −12 to −18 dBFS | −12 to −14 dBFS | −14 to −18 dBFS | −14 to −16 dBFS | −6 dBFS max |
| Mixing peak ceiling | −6 dBFS | −3 to −6 dBFS | −6 dBFS | −6 dBFS | −3 to −6 dBFS |
| Pre-master bus peak | −3 to −6 dBFS | −3 dBFS | −4 dBFS | −4 dBFS | −3 dBFS |
| True peak delivery ceiling | −1 dBTP | −1 dBTP | −1 dBTP | −1 dBTP | −1 dBTP |
| CD master sample peak | −0.5 dBFS | −0.5 dBFS | −0.5 dBFS | −0.5 dBFS | −0.5 dBFS |
| Streaming (Spotify/Apple) integrated LUFS target | −14 LUFS | −14 LUFS | −14 LUFS | −14 LUFS | −14 LUFS |
| Broadcast (EBU R128) true peak ceiling | −1 dBTP | −1 dBTP | −1 dBTP | −1 dBTP | −1 dBTP |
Values are starting-point targets; always verify with a calibrated true peak meter before delivery, and adjust for genre dynamic range requirements.
The concept of measuring the peak amplitude of an electrical audio signal predates digital recording by several decades. In the 1930s and 1940s, broadcast engineers at the BBC and NBC wrestled with the problem of transmitter overload: a program signal that exceeded the transmitter's modulation ceiling would cause splatter — interference on adjacent frequencies — and potential equipment damage. The solution was the peak programme meter (PPM), developed in the late 1930s by the BBC and standardized by 1938. The PPM used fast ballistics (attack time approximately 10 ms, slow decay of about 1.5 seconds) specifically designed to catch transient peaks that a VU meter would miss, since the VU meter's 300 ms integration time, developed in parallel by CBS and NBC engineers in the United States, was intended to reflect perceived loudness rather than instantaneous headroom usage. The two metering philosophies — peak-responding and loudness-responding — have coexisted and frequently been confused ever since.
The transition to multitrack tape recording in the 1950s and 1960s brought a new dimension to peak management. Ampex and Studer tape machines of the era had a defined nominal operating level (typically +4 dBu or 0 VU on a calibrated meter) above which the magnetic oxide coating began to saturate. Engineers like Tom Dowd at Atlantic Records and Roy Halee at Columbia Records developed intuitive practices for riding levels so that occasional peaks pushed into saturation — creating the warm harmonic density that defined recordings by Aretha Franklin and Simon & Garfunkel — without crushing the transients that gave drums and percussion their physical impact. The relationship between nominal level, peak headroom, and tape saturation was more musical than mechanical: different tape formulations (Ampex 456, BASF 911, Scotch 250) saturated at different rates and with different harmonic profiles, giving engineers a palette of controlled peak behavior.
The introduction of digital audio in professional recording contexts — beginning with the Soundstream digital recorder used on sessions including a 1978 Georg Solti/Chicago Symphony Orchestra recording, and accelerating with the Sony PCM-1600 and the commercial compact disc format standardized in 1980 — imposed a fundamentally new peak regime. Unlike tape, the 16-bit PCM format offered no graceful overload behavior: once a sample value exceeded the integer maximum (32,767 in 16-bit signed arithmetic), the system wrapped or clipped with a harsh, high-harmonic content that bore no resemblance to tape saturation. Early digital recording engineers, trained on analog practices, were surprised by how little headroom digital systems appeared to provide perceptually: a −12 dBFS peak in a 16-bit digital file used only 12 of the 16 available bits for the highest samples, reducing effective bit depth and increasing quantization noise in quiet passages. This tension between peak headroom and bit-depth efficiency was not fully resolved until 24-bit recording became standard in the mid-1990s, providing sufficient dynamic range (144 dB theoretical) that recording at conservative peak levels (−18 to −12 dBFS) imposed no practical noise floor penalty.
True peak measurement as a formal specification emerged in the 2000s alongside the EBU R128 loudness normalization standard, published by the European Broadcasting Union in 2010 following years of work by the ITU-R on recommendation BS.1770. The ITU-R BS.1770 standard, finalized in 2006 and subsequently revised, included a true peak measurement algorithm based on 4× oversampling specifically to address the intersample peak problem, which had been documented by researchers including Thomas Lund of TC Electronic and Jan Lund Abildgaard. TC Electronic's hardware meters and software tools were among the first commercially available products to implement true peak metering according to the BS.1770 specification, and the company's advocacy work — including a landmark 2006 AES paper by Thomas Lund titled "Level Practices in Digital and Analog Domain" — is widely credited with accelerating industry adoption. By 2015, Spotify, Apple Music, and Tidal had all published delivery specifications referencing true peak limits, making dBTP a term that every mastering engineer needed to internalize.
Tracking and recording: The first critical peak decision happens at the microphone preamp. Setting gain so that the loudest expected performance registers peaks between −12 dBFS and −18 dBFS on the DAW input meter gives ample headroom for unexpected dynamic moments without wasting bit depth in 24-bit recording. Engineers recording live drums — the most dynamically unpredictable acoustic source — often aim for kick and snare peaks around −12 dBFS on individual channels, knowing a rimshot or ghost-note roll can push 6–8 dB above the average performance level. Clip indicators on interface inputs should be checked after every take; digital clipping in a recorded file cannot be repaired, only obscured by downstream processing that introduces its own artifacts. The practice of deliberately overloading an analog preamp slightly before the converter — using the preamp's own saturation character as a feature — is legitimate and common, but requires that the converter input itself remains below 0 dBFS.
Mixing: During a mix, peak management operates at three levels simultaneously: individual channels, submix buses, and the master bus. Professional practice holds that individual channels should not peak above −6 dBFS before any insert processing, leaving headroom for EQ boosts and parallel effects to operate cleanly. Drum bus peaks are typically the most challenging: a dense kit with multiple close mics, overhead mics, and room mics summing to a bus can generate peaks substantially higher than any individual channel. Bus compression is the primary tool for managing drum bus peaks — not to make them louder, but to catch occasional transient excursions and glue the combined source into a controllable range. On the master bus during mixing, leaving −3 to −6 dBFS of peak headroom is essential; the mastering engineer needs space to apply their processing chain without encountering a pre-clipped signal from the first processor in the chain.
Bass and low-frequency instruments: Low-frequency content presents a particular peak management challenge because bass energy is concentrated in the range where the human ear is least sensitive to short transients. A kick drum fundamental at 60 Hz occupies a full cycle every ~16.7 ms — a duration across which a peak meter can register a sharp transient value while the perceptual loudness contribution of that single cycle is minimal. Bass guitars and synthesized bass lines with strong attack transients (plucked electric bass, square-wave synth bass) routinely generate peaks 6–10 dB above their RMS level. High-pass filtering the bass at the plugin insert input, applying a fast-attack compressor tuned specifically for transient control, and using a transient shaper to reduce attack without affecting sustain are all common techniques for taming low-frequency peak excursions without removing the perceived punch of the bass in the mix.
Mastering and delivery: The mastering stage is where peak management becomes most consequential and most technically exacting. Mastering engineers receive a mix and are responsible for delivering a file that meets the true peak specification of the target format. The chain typically ends with a true peak limiter — tools such as FabFilter Pro-L 2, Izotope Ozone Maximizer, or Weiss DS1-MK3 — set to a ceiling of −1 dBTP for streaming, −0.5 dBTP for CD, or −1 dBTP for broadcast. After limiting, a dedicated loudness/true peak metering plugin (NUGEN Audio VisLM, Izotope Insight 2, or Waves WLM Plus) verifies both integrated LUFS and true peak before export. Some engineers run the final master through a free online tool such as the Loudness Penalty Analyzer to simulate how streaming platforms will normalize the file, confirming that the combination of integrated loudness and peak behavior will translate as intended to the listener.
One email a week. The techniques behind the terms — curated by working producers, not algorithms.
Abstract knowledge becomes practical when you can hear it in music you know. These tracks demonstrate peak used intentionally, at specific moments, for specific purposes.
The opening four bars feature an isolated bass-heavy synth hit and kick that read extremely high on a peak meter relative to their short duration — a deliberate use of peak-heavy, LUFs-efficient programming. Mike Will Made-It's production on this record is a textbook case of using tight peak limiting (the master sounds heavily processed with a ceiling near −1 dBFS) to allow the track to compete loudly on streaming while passing normalization at −14 LUFS without subjective volume loss. Listen on headphones: the kick's initial transient is noticeably hard-limited, giving it a punchy, almost clipped character that suits the aggressive aesthetic.
This track was mastered with significant attention to true peak compliance for Apple Music's Dolby Atmos delivery alongside the standard stereo master. The minimalist arrangement — sparse bass hits, vocal, and percussion — means peaks are isolated and easily audible as discrete events rather than continuous density. The bass drop at 0:43 is an excellent example of a controlled low-frequency peak: the sub hits hard on the meter but the mastering has preserved the transient cleanly, demonstrating how −1 dBTP limiting applied thoughtfully preserves punch rather than crushing it.
Mastered by Bob Ludwig at Gateway Mastering, this track is frequently cited as an example of commercially loud mastering that retains dynamic integrity — an achievement of careful peak management rather than brute limiting. The guitar, bass, and drum interplay produces peaks that are spread across the frequency spectrum and never concentrate on a single transient, making the limiter's job easier. Check the snare in the verse: the peak is sharp and clean, indicating the limiting ceiling was set high enough (approximately −0.5 dBFS sample peak) that individual snare transients pass through without being obviously rounded.
Recorded and mixed in the pre-streaming era when the Loudness War incentivized aggressive peak limiting, the 1994 master of this track shows deliberate use of near-0 dBFS peak levels throughout its dense industrial arrangement. This is historically instructive: the track sounds loud for its era precisely because peaks are allowed to approach 0 dBFS on the master, creating a dense, saturated quality. Comparing the original master to a modern remaster reveals how the same mix can be presented with more headroom and recover dynamic detail. A useful A/B exercise for understanding how peak ceiling choices affect perceived character.
Sample peak measures the absolute maximum value of individual digital samples in the audio stream, with no oversampling or interpolation. It is the most common form of peak metering and is displayed in dBFS on every DAW channel strip. Sample peak metering is fast and computationally trivial, but it does not detect intersample peaks that occur during DAC reconstruction, making it insufficient as a sole delivery metering standard.
True peak metering applies oversampling (typically 4× to 8× the native sample rate) to reconstruct the continuous-time representation of the digital signal and detect peaks that occur between sample points. Mandated by EBU R128, ITU-R BS.1770-4, and all major streaming platform delivery specifications. True peak values are expressed in dBTP (decibels true peak) and are always equal to or higher than sample peak values on the same signal; a signal with a sample peak of −0.5 dBFS may measure +1.0 dBTP on a true peak meter after lossy re-encoding.
The Peak Programme Meter, standardized by the BBC in the 1930s, uses fast attack ballistics (~10 ms integration) and slow decay (~1.5 seconds) to catch transient peaks that VU meters miss while remaining readable during rapid program material. PPM scales vary by standard: the EBU PPM scale labels 0 dBu as PPM 4, with the permitted maximum at PPM 6 (+8 dBu). PPMs are primarily an analog and broadcast standard but remain in use in European broadcast facilities and are emulated by software metering plugins for engineers working with broadcast delivery specifications.
A subset of true peak measurement focused specifically on identifying peaks that occur between digital samples, which are generated by DAC reconstruction filtering and amplified by lossy codec encoding. ISP detection is implemented in modern true peak limiters as part of their gain reduction algorithm: rather than limiting based on sample values alone, the limiter reconstructs the inter-sample waveform before applying gain reduction, ensuring that neither sample peaks nor intersample peaks exceed the specified ceiling. This eliminates the category of post-encoding clipping that affected CD-era masters delivered at −0.1 dBFS sample peak.
The clip indicator is a binary metering element — an LED or screen segment — that activates the moment a signal reaches or exceeds 0 dBFS. Unlike continuous meters, the clip indicator is latching by design: once triggered, it remains illuminated until manually reset, ensuring that even a single clipped sample during a long performance is not missed. Modern interfaces such as the Universal Audio Apollo series extend this concept with sample-accurate clip logging in the console software, allowing engineers to identify exactly when and on which channel clipping occurred during a session.
These MPW articles put peak into practice — specific techniques, real tools, and applied workflows.