← All posts ·

RMS vs. Peak Compression: What the Detector Actually Hears

Rack compressor in an audio test room

Peak and RMS describe what the compressor's detector measures, not how hard it squeezes. A peak detector reacts to the highest instantaneous level in the signal. An RMS detector reacts to the average energy over a short window, which tracks perceived loudness more closely. Peak is the safer pick for catching transients. RMS is the smoother pick for sustained material.

The rest of this article covers what that means at the knobs, on which sources to use each mode, and where each one fails. We are a small design studio and we think about detectors a lot, so we will stay practical.

TL;DR

Table of Contents

What each detector measures

Every compressor has two parts: a level detector and a gain-changing stage. The AES paper on attack and release constants describes the core of a compressor as a level detector plus an amplifier with variable gain. The detector turns the incoming signal into a control value. The gain stage then does what that value says.

Peak and RMS are two ways of building that detector.

Peak detection triggers gain reduction from the instantaneous maximum amplitude of the signal, which makes it very responsive to rapid transients. When a spike crosses the threshold, the compressor acts.

RMS detection triggers gain reduction from the average signal energy over a defined time window. The same source notes this more closely mimics how people perceive loudness.

The math explains the name. The RMS value is the square root of the mean of the squared instantaneous values of the waveform. Squaring emphasizes larger values, averaging smooths them over time, and the square root brings the result back to signal units.

Why the ratio between the two moves around

For a pure sine wave, RMS is about 0.707 times the peak value. That is roughly 3 dB below the peak. Music does not behave like a sine wave. According to the same Juniata text, complex musical signals have no fixed ratio, so an RMS-calibrated compressor reacts differently to material with a high peak-to-average ratio than to material with a low one.

Here is an illustrative case, not a measurement. Picture a dry snare hit and a sustained synth pad, both peaking at the same level on your meter. The pad has a lot of energy between its peaks, so its RMS value sits close to its peak. The snare has a sharp spike and then quickly falls away, so its RMS value sits far below its peak. Set the same threshold on an RMS detector and the pad will trigger far more gain reduction than the snare, even though the meters show the same maximum.

Equal-peak snare and pad detector response comparison

A common misconception: RMS is not slow peak

People often describe RMS as the slow mode and peak as the fast one. That is partly true in practice, but it hides the real difference. RMS changes the quantity being measured: it follows power, not height. A peak detector with a long release is still a peak detector. It still reacts to the tallest point, and it still holds on to a spike.

The Cubase compressor reference shows this clearly. It offers an analysis control that runs from pure peak at 0 to pure RMS at 100, with mixtures in between. Blending the two is a legitimate design choice, because each one solves a different problem.

How to choose: source by source

Practitioners generally use peak detection on transient-heavy sources such as drums and percussion, where control over individual peaks matters. RMS is typically favored for sustained sounds such as vocals, pads and mix buses, where a response that follows perceived loudness is the goal.

Source Starting detector Why
Kick, snare, percussion Peak Fast transients need to be caught before they overshoot
Lead vocal RMS, or a blend Follows how loud the phrase feels, so level changes sound even
Bass guitar RMS or blend Sustained notes carry the energy, with some plucked transients
Pads, strings, keys RMS Little transient content, so averaging works well
Mix bus RMS, light peak help Glue without reacting to every stray spike
Final protection Peak (limiter) Prevents overs that averaging can miss

Treat that table as a starting point. Your source, tempo and taste can overrule it.

Drums: peak for control

On a snare, the first few milliseconds carry the attack and most of the peak level. A peak detector sees that spike right away and can bring it down. An RMS detector may average the spike with the quiet moments around it, so the transient can pass mostly intact. That can be exactly what you want if the goal is to keep the snap, but it is not control.

If you want to shave an over-aggressive transient, choose peak. If you want the body to sit more evenly while the snap stays, choose RMS and listen to what survives.

Vocals and buses: RMS for feel

One write-up from a university compact mixer reference guide notes that users often find peak mode on a full mix sounds chunky or distorted, while RMS keeps the feel of the original performance by averaging over a wider window. That matches what many engineers hear on a stereo bus. Peak detection on a dense mix reacts to the kick, then to the snare, then to the vocal, and every one of those reactions is audible.

On a lead vocal, RMS follows how the phrase swells and falls, so gain changes feel like a hand riding a fader rather than a clamp on each consonant.

What RMS lets through

This is the part many engineers wish they had learned earlier. An RMS detector can let potentially clipping transients pass because a brief spike barely moves the average inside its window. The gain reduction you hear is based on the average, so the peak that matters for headroom can stay almost where it was.

That has real consequences:

The practical fix is to pair stages. Let an RMS-style compressor shape the body and tone, then use a peak-sensitive limiter or fast peak compressor to catch what the average missed. Watch the peak and RMS readings on your meter together, not either one alone.

What peak detection costs you

Peak detection is not a free upgrade. Because it reacts to individual spikes, it can respond to moments the ear does not register as loud. A very short peak may be reduced even though it contributes little to perceived loudness. On sustained material, that can produce gain changes that sound busier than the music.

It also interacts strongly with the attack and release settings. With very fast attack, a peak detector may react within a few milliseconds and soften the very edge that gives a drum its character. Set the attack slightly slower and some of the transient survives before gain reduction begins. That trade is the heart of drum compression.

Where pumping comes from

When a mix breathes in a way you do not like, check the dynamics processing first: threshold, ratio, attack and release. Pumping is the audible result of gain reduction moving up and down with the program material, and that motion is governed by those dynamics settings. A release that recovers too fast on a loud, sustained low end will lift the level between hits. A release that is too slow can hold the gain down and flatten the next phrase.

Changing the detector from peak to RMS changes when and how much the gain moves, so it can reduce or exaggerate the effect. But the cause sits in the compression curve and the timing, so adjust those first.

A setup workflow you can repeat

This workflow works with any compressor that offers a peak/RMS choice or a blend.

  1. Set the detector to match the job. Peak for transient control, RMS for evenness and feel.
  2. Start with a moderate ratio. Something like 2:1 to 4:1 is a common starting zone for general work, then adjust by ear.
  3. Set the attack. Slower lets transients through, faster clamps them. On RMS, remember the detector already smooths the spike.
  4. Set the release. Watch the gain reduction meter. It should return to rest before the next important moment, or you will hear it hold the level down.
  5. Lower the threshold until you see 2 to 4 dB of gain reduction. Then compare bypassed and active at matched loudness, not at different levels.
  6. Check the peaks. Look at the output peak meter. If overs remain, add a peak stage afterward.

Matching loudness in step 5 is essential. A louder version nearly always sounds better, so compare at equal levels or you will pick whichever is simply louder.

Combining detectors with parallel compression

Parallel compression gets around the peak versus RMS decision. You blend a heavily compressed copy with the dry signal. The dry path keeps the transients, and the compressed path adds density. Because the detector on the compressed path only has to support the body, RMS often suits it well.

We cover this approach in detail in our post on New York compression and how to blend it. If the low end and the highs need different treatment, our guide to multiband parallel compression with per-band blends shows how to compress the low band and the top band differently.

If you push the signal with a little saturation first, remember that adding harmonics changes what the detector sees. We tested that in our 2 to 4 dB test on saturation before compression. One caution there: saturation can change the spectrum and the peaks in ways that make compressed levels harder to compare, so do not assume it makes a track louder for free.

If you are tracking, detector choice also interacts with latency. Look-ahead lets a compressor see a spike before it arrives, at the cost of delay, which we discuss in keeping monitoring responsive with 1 to 2 ms of lookahead.

Our perspective: ask what each band needs

We design plugins around one observation: a full-range detector makes one decision for the whole signal. A kick and a vocal in the same bus drive the same detector, and the loudest low-frequency event decides what happens to everything else. That is why a single peak or RMS choice can feel like a compromise on a full mix.

Our answer is to split the decision by frequency. Low bands often have sustained energy, and RMS-style behavior tends to suit them. Upper bands carry the attacks, the consonants and the cymbal edges, and a faster, peak-aware response can suit those. When each band gets its own detector behavior and its own blend, you can control a problem region without dragging the rest of the mix along.

Our perspective: ask what each band needs — overview diagram

We also keep two cautions in mind. First, any filter that splits the signal shapes phase, and even a minimum-phase filter still has group delay, so band splitting is never free of timing consequences. Second, a detector cannot fix what the settings cause: if the release is wrong, changing from peak to RMS only hides it.

If you take one idea from us, make it this: choose the detector by what you want the compressor to notice. Spikes, or sustained loudness.

— Kai

Doing it with a lane-based approach

If you want to try this per-frequency approach in one insert, ToneLab is one way to do it: parallel effect lanes, each targeted at its own frequency region. It is a one-time license at 73.99 USD, and there is a free demo so you can test the idea on your own session before buying.

Common mistakes

FAQ

Is RMS or peak compression better?

Neither is better in general. Peak detection is stronger for controlling transients on drums and percussion. RMS detection is usually more natural on vocals, pads and mix buses. Many engineers use both in a chain.

What does RMS mean on a compressor?

It tells the compressor to react to the averaged energy of the signal over a short window, instead of the single highest instant. RMS stands for root mean square, which is the square root of the mean of the squared values of the waveform. It tracks perceived loudness more closely than peak does.

Why does my peak level stay the same after RMS compression?

Because the RMS detector responds to average energy, short spikes can pass with little reduction. The average level drops while the highest peaks stay about where they were. If you need those peaks lower, add a peak-based compressor or limiter afterward.

Can I use RMS and peak at the same time?

Yes. Some compressors offer a blend control between the two, and you can also chain two compressors. Use RMS for shape and feel, then peak for protection.

Sources