What are dynamic range and crest factor?

Dynamic range and crest factor in plain English

Every sound has loud moments and quiet ones. Dynamic range is the distance between them, in decibels, and it's what people mean when they say a recording sounds dynamic or squashed. A string quartet that goes from a whisper to full force has a wide dynamic range. A pad that holds one chord at one level has a narrow one.

The trouble with the phrase is that it covers changes on very different time scales. The front of a snare hit lasts a few milliseconds; a quiet verse can last thirty seconds. So engineers measure dynamic range in more than one way, and the one you'll meet most often while mixing is the crest factor.

Crest factor compares two readings of the same stretch of sound. One is the peak, the single highest point the waveform reaches. The other is the RMS level, a kind of average that follows how much energy the sound carries. Subtract the RMS level from the peak and you have the crest factor. A big number means short spikes standing well above the body of the sound; a small one means the sound sits close to its peak most of the time.

None of this says how loud a track is. A loud master and a quiet one can have the same crest factor, and turning either of them up or down won't change it. Loudness is measured in LUFS, which the LUFS lesson covers. This page is about the shape of the level, and what squeezes it.

Reading the lab

The lab plays a loop through a limiter and a gain control and measures everything that comes out. The screen to watch here is the one labelled Crest factor. Each column is one version of the loop: the left one is the original and the right one is your master. The amber block is the RMS level, the pink line on top of it is the highest sample peak, and the number between them is the crest factor in dB.

At the bottom of the Right now panel, Crest gives the same figure as a number, with the original's value under it. Next to it, LRA is the loudness range, which measures the slower kind of dynamic range from one section of the music to the next. Both change when you pick another sound or push the limiter.

Gain is plain volume. The Limiter knob pushes the loop into a limiter whose ceiling is −1 dBTP, which holds the peaks down while the rest comes up. The Streaming buttons play both versions at the level Spotify, YouTube or Apple Music would, so you can hear what a smaller crest factor costs once the loudness is equal. The lab sets every loop so its highest sample is close to 0 dBFS, and you can load your own sound with the Your sound button.

What is crest factor?

Crest factor is the peak level of a signal divided by its RMS level. Written in decibels, the division becomes a subtraction:

crest factor (dB) = peak level (dBFS) − RMS level (dBFS)

Suppose a drum bus peaks at −2 dBFS and its RMS level over the same eight bars is −18 dBFS. Its crest factor is 16 dB. If a compressor and its makeup gain leave the RMS level at −18 dBFS with the peaks now at −5 dBFS, the crest factor is 13 dB. The bus reads the same on an RMS meter and its peaks are 3 dB further from the ceiling.

You'll sometimes see crest factor written as a plain ratio. A sine wave's peak is 1.414 times its RMS value, which is the square root of 2, and 20 × log10(1.414) is 3.0 dB. A square wave sits at its peak all the time, so its ratio is 1 and its crest factor is 0 dB. Triangle and sawtooth waves measure 1.732, or 4.8 dB. Those figures are fixed by the shape of the wave, which makes them handy for checking a meter.

Because crest factor is a difference between two levels, gain can't change it. Turn a sound down 10 dB and the peak and the RMS level both fall by 10, leaving the gap where it was. Only processing that treats the loud and quiet parts differently can move it, which is why compressors, limiters, clippers and saturation all change crest factor and a fader doesn't.

Some RMS meters follow a convention from the AES17 standard and add 3 dB to every reading, so that a full-scale sine reads 0 dBFS RMS. On those meters every crest factor comes out 3 dB smaller. Check which kind you have before comparing numbers with someone else's.

Crest factor of common sounds

Sounds with sharp attacks and gaps between them have the highest crest factors, since the gaps pull the average down and the attacks push the peak up. Sounds that hold one level sit at the bottom. These are the readings from the loops in the lab, measured across the whole loop with nothing applied:

Every loop in the table peaks close to full scale, yet the integrated loudness runs from −21.8 LUFS for the acoustic guitar to −2.5 for the sine. Crest factor is most of the reason. With the peak fixed at the top of the scale, a sound with a 25 dB crest factor has its average level 25 dB down, and the average is where most of the loudness comes from.

This is also why drums decide how much headroom a mix needs. A kick and snare with an 18 dB crest factor hit the ceiling long before the mix sounds loud, while a pad with a low crest factor can sit at a high average level and barely move the peak meter. The headroom lesson follows that through the recording and mixing stages.

What does dynamic range mean?

The phrase gets used for different measurements, so it helps to know which one a person means.

The dynamic range of a system

For equipment and file formats, dynamic range is the distance from the noise floor up to the loudest level the system can handle without clipping. A 16-bit file has a theoretical range of about 96 dB and a 24-bit file about 144 dB, roughly 6 dB for every bit. Converters and preamps are specified the same way. This kind of dynamic range belongs to the container, and it says nothing about the music inside it. The sample rate and bit depth lesson covers it.

Short-term dynamics: crest factor

Inside the music, the fastest movement is the jump from a transient to the body of a note. Crest factor measures that, and engineers sometimes call it micro-dynamics. It's what gives drums their snap and a plucked guitar its edge, and it's the first thing a limiter takes away.

Long-term dynamics: loudness range

The slower movement, from a quiet intro to a loud chorus, is measured by loudness range, or LRA. The EBU defined it in a document called Tech 3342. A meter takes the short-term loudness of every three-second stretch, throws away the very quietest moments, and reports the spread between the 10th and 95th percentile of what's left, in LU. A heavily compressed pop song might measure only a few LU, while orchestral recordings often measure well over 10.

The two don't have to agree. In the lab, the drum loop and the chords have exactly the same crest factor, 17.6 dB. The drum loop has a loudness range of 0.7 LU because every bar is as loud as the last, while the chords measure 5.4 LU because they swell and fade over several seconds.

The vocal has plenty of both, with a crest factor of 19.2 dB and a loudness range of 8.2 LU, which is why vocals are usually the first thing in a mix to get compressed.

Ways to measure dynamic range

Each measurement answers a slightly different question, and most loudness meters show more than one of them.

PLR, the peak to loudness ratio, is the one to know for masters. It's a crest factor taken with a loudness meter, worked out by subtracting the integrated loudness in LUFS from the true peak in dBTP, so a bigger number means more room between the loudest peak and the average level of the whole track. The lab's full mix peaks at −0.2 dBTP and reads −14.1 LUFS, so its PLR is 13.9 LU. After 12 dB of limiting it peaks at −1.0 dBTP and reads −9.6 LUFS, a PLR of 8.6 LU.

The stretch you measure over matters more than beginners expect. Measure the crest factor of one snare hit and you'll get a very different number from the same snare across a whole song, because the song includes the gaps between hits. When you compare two crest factors, make sure they were measured over the same kind of stretch.

Compression, limiting and crest factor

A limiter holds the peaks at its ceiling and lets everything underneath come up, so its whole job is to shrink crest factor. In the lab, 6 dB of push brings the full mix's crest factor down from 16.2 to 12.2 dB and makes it 3.2 LU louder.

Each extra decibel of push buys less. This is what the lab measures on the full mix as the push goes up:

The first 6 dB of push takes 4 dB off the crest factor. The next 12 dB takes less than 2 dB more, because by then the limiter is turning the body of every hit down along with its peak, and the average level falls with it. Past that point, more limiting mostly adds distortion and flattens the drums.

How much a limiter takes depends on where a sound's energy is. With 9 dB of push, the chords' crest factor falls from 17.6 to 10.4 dB and the vocal's from 19.2 to 11.6 dB, and both get more than 6 LU louder. The drum loop only drops from 17.6 to 15.0 dB and gets 2.6 LU louder, because most of its energy is in the very hits the limiter turns down.

A compressor does the same kind of thing more gently, and its attack setting decides which kind of dynamics it works on. A fast attack catches the transients and lowers the crest factor. A slow attack lets the transients through and works on whole notes and phrases, so it leaves more of the crest factor, and with a slow release it can narrow the loudness range. The compression lesson lets you hear both. A clipper lowers crest factor instantly by cutting off the tops of the peaks, as the figure above shows with a sine.

Going the other way is possible too. A transient shaper turns the start of each hit up and raises the crest factor, and an expander makes quiet parts quieter, which widens the dynamic range. The transient shaping lesson covers the first of those.

Crest factor after streaming normalisation

Streaming services measure the integrated loudness of every track and turn it to their own level, which is −14 LUFS on Spotify. The loudness a limiter adds is taken away again on playback, and what's left is the crest factor. At equal loudness, the master with the lower crest factor has lower peaks.

In the lab's last step the full mix has been pushed 12 dB into the limiter. Spotify turns it down 4.4 dB to −14.0 LUFS, where the original already plays, so the two are equally loud. The original's peaks reach −0.2 dBFS and the limited master's reach −5.4 dBFS. Press A/B and listen to the kick and snare: in the limited version they're 5 dB smaller.

PLR tells you in advance how this will go. If the true peak can't go above −1 dBTP and a service plays tracks at −14 LUFS, a track needs a PLR of 13 LU or less to reach that level with its peaks intact. A mix with a PLR of 16 LU can only get to −17 LUFS before its peaks reach −1 dBTP, and services that turn quiet tracks up, as Spotify does, stop there. The streaming loudness lesson covers each service's rules.

How much dynamic range should a mix have?

There's no target crest factor for a genre. What you want is a mix where the parts meant to hit still hit, and the arithmetic above gives you some reference points for that.

A mix with no limiting on it often has a crest factor in the teens, like the lab's 16.2 dB. A master at −14 LUFS with its true peak at −1 dBTP has a PLR of 13 LU, which keeps most of that. A master at −9 LUFS with the same peak has a PLR of 8 LU, and on Spotify it will play at the same level as the −14 LUFS master with 5 dB less between its peaks and its average.

For music that will mostly be streamed, a PLR of about 9 or 10 LU is a sensible place to stop limiting, because past there you're giving up transients for loudness that normalisation takes back. Some styles are meant to sound squashed, and there the heavy limiting is part of the sound.

For loudness range, think about where the music will be heard. Eight LU in a vocal is fine on headphones, but in a car or a noisy room the quiet lines drop under the road noise, which is one reason vocals are compressed on most releases. Film and classical music can keep much wider ranges because people tend to listen to them somewhere quieter.

How to measure crest factor and dynamic range, step by step

1. Put a meter on the masterUse one that shows true peak, integrated and short-term loudness and loudness range. Youlean Loudness Meter has a free version that shows all four, and PLR too, and many DAWs have a loudness meter of their own.
2. Play the whole track from the startIntegrated loudness and LRA need the full length to be accurate. Reset the meter first so it isn't counting anything you played earlier.
3. Work out the PLRSubtract the integrated loudness from the highest true peak. A track at −11 LUFS with peaks at −1 dBTP has a PLR of 10 LU.
4. Read the loudness rangeThis tells you how much the track moves between sections. Compare it with a reference track you like in the same genre.
5. Check a single part with peak and RMSFor one channel, loop its busiest section and read peak and RMS on the same meter. The difference is the crest factor of that part.
6. Measure before and after each processorBypass a compressor or limiter and read again. If the crest factor drops by several dB and you can't hear an improvement at matched loudness, back it off.

Common dynamic range mistakes

Confusing dynamic range with headroom

Headroom is the space between your loudest peak and 0 dBFS. Dynamic range is the space between the loud and quiet parts. A mix can peak at 0 dBFS with no headroom left and still have a crest factor of 18 dB, and a mix peaking at −10 dBFS can be squashed flat.

Expecting volume to change the dynamics

Turning a track down doesn't make it more dynamic, and normalising it doesn't make it less. Only processing changes the gap between peaks and average.

Losing crest factor one channel at a time

A compressor on every drum, then one on the drum bus, then a clipper and a limiter on the master: each takes a little, and the total can leave the drums with no snap before mastering has started. Measure the crest factor of your drum bus with and without its processing, and keep the reduction to what you can hear doing a job.

Comparing at different loudness

The limited version is louder, and louder usually sounds better for the first few seconds. Use the Streaming buttons in the lab, or turn the louder version down by hand, and compare the two at the same LUFS.

Comparing numbers from different meters

An AES17 RMS meter reads 3 dB higher than a plain one, DR scores use their own method, and a crest factor taken over one bar won't match one taken over the whole song. Only compare readings that were made the same way.

Train your ears

The quiz under the lab plays two versions of a loop with the screens covered and asks which has the higher integrated loudness. In most rounds one version has been through the limiter, so it has a lower crest factor and lower peaks, and it can be the louder one even though a peak meter would say otherwise. Eight rounds take a few minutes, and your best score is saved in your browser.

Questions people ask

Keep going