Audio meters, made simple
Peak. True Peak.
RMS. LUFS.
Four meters. One easy picture.
Peak checks the dots. True peak checks the wave between them. RMS checks how much sound keeps going. LUFS estimates how loud it feels.
FREE GUIDE / NO SIGN-UP / 5 MIN READ
SAME SOUND. FOUR DIFFERENT QUESTIONS.
TAP A VIEW TO EXPLOREPeak · the highest dot
−2.50 dBFS
True peak · ideal wave
+0.51 dBTP
LUFS · how loud it feels
Needs a longer view
Think of the dots as photos of a moving wave. A photo can miss the very top. The line shows what happens between photos. The shading helps you picture how much sound keeps going. Each meter looks at a different part of that story.
Sample peak
Which dot is highest?
A computer stores sound as lots of little snapshots, called samples. These are the dots. Peak finds the dot furthest from the middle line. It is useful for checking a sudden hit, like a kick or clap.
True peak
Does the wave go higher?
Playback rebuilds a continuous wave from the samples. That wave can go higher than any dot. A true-peak meter estimates its biggest peak, including peaks between samples. Both upward and downward swings count.
RMS
How much sound keeps going?
Think of a short beep versus a held beep. Both can reach the same height, but one keeps going. RMS describes average signal level across a chosen stretch of time. It depends on both level and duration, not simply how long a sound lasts.
LUFS
How loud does it feel?
Our ears do not treat all pitches equally. LUFS measures average energy with rules that better match how we hear. It helps compare the loudness of a section or a whole track.
Look between the samples
A photo can miss
the highest bounce.
Imagine taking a few photos of a bouncing ball. None might catch the exact moment it reaches the top. The dots are like those photos. The curve shows the movement between them.
Try it: move the snapshots.
Move the slider. The wave stays exactly the same, but the dots land in different places. Watch how the highest dot can miss the top of the wave.
In the middle, the dots miss the tops most. At either end, they catch them.
At 45°, sample peak is −2.50 dBFS while the reconstructed peak is +0.51 dBTP.
The dots stay below the ceiling, but the wave goes above it. This is why a normal peak meter can look safe while a true-peak meter shows a warning.
The pink lines mark the positive and negative full-scale limits. Both mean 0 dBFS in magnitude. Crossing them can cause clipping later; it does not prove that distortion is already audible.
Remember it this way: peak checks the photos; true peak checks the whole movement. The real curve can rise between the dots. Drawing straight lines from dot to dot would hide that extra height.
The photo analogy has a limit: properly sampled, band-limited audio contains enough information to reconstruct the wave between samples. The sample-peak meter misses the peak; the recording has not necessarily lost it.
How this diagram works (and what it does not claim)
This is an ideal sine wave with four equally spaced samples per cycle. It represents a repeating signal continuing beyond the picture; at a 48 kHz sample rate its frequency would be 12 kHz. The slider changes where sampling begins, not the wave itself. Quantisation is omitted.
The wave’s peak amplitude is 1.06, relative to full scale of 1, so its ideal true peak is 20 × log₁₀(1.06) = +0.506 dBTP. Halfway along the slider, sample peak is 1.06 ÷ √2, or −2.504 dBFS. At either end the dots hit the extrema. The values are calculated from this known sine, not from a standards-compliant meter running in the page.
This illustration remains unclipped. At some slider positions the samples themselves exceed full scale; they would need attenuation to be stored without clipping in an ordinary fixed-point file. A real clipped signal would have a different shape.
Real true-peak meters use filtered oversampling to estimate the waveform’s peaks, with finite accuracy. The estimate does not guarantee the exact output of every converter or codec. Lossy encoding and other processing can change peaks: check the final delivered file where possible.
A quick beep versus a held beep
Same peak.
Different average level.
Play the same note at the same level: first briefly, then hold it. Both reach the same peak. Across the same stretch of time, including the quiet part, the held note has a higher RMS level. We are comparing average signal level here, not claiming a particular LUFS result.
Both pictures use the same time span and peak height. The waves are illustrations; no audio is playing.
A / QUICK BEEP
Beep. Then quiet.
It reaches the top, then stops. Its average level over this whole picture is lower.
B / HELD BEEP
Beeeeeeep.
It reaches the same top, but keeps going. Its average level over this whole picture is higher.
“Energy inside the wave” is a memory aid, not the calculation. RMS is an average level related to signal energy, not energy itself or the coloured area. LUFS adds hearing-related weighting. For whole-song LUFS, very quiet sections can be left out, so more silence does not necessarily mean a lower reading.
The extra detail: how loudness is measured
LUFS applies K-weighting, then combines mean-square values using channel weights and a logarithmic scale. K-weighting is a practical approximation, not an exact equal-loudness curve. Integrated LUFS discards sufficiently quiet blocks (gating). The momentary and short-term views in EBU Mode are not gated. LUFS means Loudness Units relative to Full Scale; LKFS names the same standardised loudness unit.
That means a long silence does not simply keep pulling integrated LUFS down. These pictures explain the idea of energy; they are not measured LUFS results. Your speaker volume, listening distance and room also affect what you hear, without changing the file’s LUFS. The beeps above illustrate RMS only; BS.1770 cautions against using its algorithm to predict the perceived loudness of pure tones.
The easiest way to separate them
RMS measures average level.
LUFS considers our ears.
Two sounds can have the same average signal level and still feel different in loudness. Our ears are more sensitive to some frequencies than others.
RMS: “How much is going on?”
Think of the short beep and held beep above. Peak sees the biggest instant. RMS considers the whole measurement window, including quiet parts. Both the height of the wave and how long it stays large matter. A brief, much louder sound can still have a higher RMS than a quiet sustained one.
LUFS: “How loud does it feel?”
Now think about a bass-heavy passage and a bright, midrange-heavy passage. They can have the same RMS reading without seeming equally loud. LUFS uses frequency weighting to make a useful loudness estimate for music and speech.
LUFS is not an exact model of every listener. It is especially unreliable as a prediction of how loud isolated pure tones feel.
What does RMS actually stand for?
Root mean square. Square each sample value, average the results over the chosen window, then take the square root. Squaring stops the wave’s positive and negative sides cancelling each other out. RMS is an amplitude quantity; mean-square is proportional to average power under the same conditions. RMS windows and channel handling vary between meters.
A sine wave with a peak amplitude of 1 has an RMS amplitude of about 0.707. On an amplitude-referenced scale, that is 0 dBFS peak and −3.01 dBFS RMS. Meters using a full-scale-sine RMS reference show that same sine at 0 dBFS RMS. This 3.01 dB display offset is a calibration convention, not extra signal level. Check the reference before comparing meters.
The peak-to-RMS amplitude ratio is crest factor. In decibels it is the difference between peak and RMS levels measured over the same span, after correcting any RMS display offset. A bigger ratio means peaks stand further above the RMS level. That can indicate stronger transients, but silence and waveform shape also affect it. It does not tell you whether the mix sounds good.
A glance, a phrase, or the whole song
LUFS has three views.
Momentary
A quick glance at loudness right now. It responds to short changes.
Short-term
A few seconds at a time. Useful for seeing how a drop compares with a breakdown.
Integrated
The overall reading for what you have played, with sufficiently quiet sections excluded. Reset the meter, then play the whole track to measure the full song.
The minus signs can look backwards: −9 LUFS is a higher measured loudness than −14 LUFS. For negative readings, less negative means higher. LUFS has no clipping ceiling at zero. Always compare the same view and channel layout, at the same point in the signal chain.
Bring it back to your mix
Which meter should I watch?
| Measurement | Unit | What it helps you check |
|---|---|---|
| Sample peak | dBFS | “How high are my biggest hits?” |
| True peak | dBTP | “Does the wave sneak above the dots?” |
| RMS level | dBFS RMS* | “How much signal level keeps going?” |
| Loudness | LUFS | “How loud is this passage or song likely to feel?” |
*RMS meter calibration varies. Compare like-for-like meters and time windows; some RMS meters use a sine-referenced scale with a 3.01 dB offset.
A simple way to use them
While mixing, use peak to watch the biggest hits and leave room for them. Use RMS to understand how sustained the level is. Use LUFS to compare loudness. Before sending out the finished track, check true peak on the export. Then listen: a good number is not the same as a good mix.
Do not chase a number just because it looks right. Making a track louder can also take away punch. When comparing two versions, turn them to a similar listening volume so the louder one does not win automatically.
Starting points, not a score to beat
Pre-master and master:
different jobs, flexible targets.
Think of a pre-master as a finished meal before the final seasoning. Mastering is that last adjustment for how it will be served. There is no single amount of seasoning that suits every dish, and no single loudness number that suits every track.
PRE-MASTER / THE MIX YOU HAND OVER
Leave room. Keep the punch.
A pre-master here means the mix you send for mastering. A convenient starting suggestion is to let the loudest sample peaks sit roughly between −6 and −3 dBFS. This leaves a few decibels of headroom: space below the digital ceiling.
That is a working habit, not a technical requirement. An unclipped mix outside that range can be perfectly usable. Ask your mastering engineer what they want. There is no universal pre-master LUFS or RMS target.
Do not compress or limit the mix just to land on a number. If a limiter is only there to make your rough mix loud, discuss supplying a version without it, plus the loud reference. Keep intentional mix processing that defines the sound, and explain it to the engineer.
MASTER / THE FINISHED RELEASE
Choose the sound, then check the numbers.
For a listening experiment, you could compare versions around −14, −11 and −8 LUFS integrated. These are example comparison points, not recommended release targets, a required range, or a claim about what all trance masters measure.
A spacious track may work quieter; a dense club track may suit a louder approach. Stop pushing if the kick loses impact, the bass distorts or the mix feels tiring. Compare at matched listening volume, so extra loudness does not win by itself.
For a general starting point, consider a true-peak ceiling around −1 dBTP or lower, then check the destination’s guidance and the exported file. More margin may be appropriate for lossy encoding. A ceiling setting is not a guarantee that every later conversion will stay below it.
Turning a mix down creates level headroom, not lost dynamics. Clean gain reduction before export can lower peaks without changing the balance. It cannot undo clipping, distortion or flattened transients already caused earlier in the chain.
−14 LUFS is not a universal mastering rule.
A service’s playback normalisation level and your artistic mastering target are different decisions. As one example, Spotify describes −14 LUFS normalisation for its Normal setting and gives its own mastering recommendations. Other settings and destinations differ. Louder masters may simply be turned down during playback.
For context, Spotify’s current guidance recommends a maximum of −1 dBTP, and below −2 dBTP for masters louder than −14 LUFS integrated. That is destination-specific advice, not a universal law. Check it again when delivering. iZotope’s mastering preparation advice likewise notes that engineers’ headroom requests vary.
What still matters: avoid unintended clipping, preserve the sound you want, and meet any actual delivery specification you have agreed to. A label or mastering engineer’s explicit requirement takes precedence over the broad suggestions here. Compare like-for-like sections and meter modes: a drop’s short-term LUFS is not the same measurement as a whole track’s integrated LUFS. There is no universal finished-master RMS target either.
The questions that usually follow
Still wondering?
Is true peak always higher than sample peak?
For ideal reconstruction, true peak is at least as high as sample peak: the reconstructed wave passes through its samples. If a dot lands at the wave’s largest peak, the two match. Real true-peak meters approximate that reconstruction, so small differences and under-reads are possible.
What does “0 dBFS” mean?
Think of it as the digital ceiling for sample values in a normal fixed-point audio file. A reading of −6 dBFS is below that ceiling; 0 dBFS reaches it. True peak can reveal that the wave between samples goes higher. Above-ceiling levels can distort at an output or conversion stage, even if a floating-point mixing session can hold them internally.
Is LUFS the same as RMS?
No. RMS tells you about average signal level. LUFS adjusts the energy measurement to better match how we hear. Think “how much signal?” for RMS and “how loud does it feel?” for LUFS.
Does every master need the same LUFS?
No. Different music needs different amounts of punch and space. Check the requirements of whoever receives your track, but do not squash the life out of a mix just to hit a number. A streaming service’s playback level is not automatically a target you must master to.