Dynamic Range

I’ve been investigating the dynamic range of some music, using this plugin for the foobar2000 music player: http://www.pleasurizemusic.com/de/free-downloads. This tool is useful for analysing my own music, for improving my mixes. I also tested the albums that I use for reference, and then a few more.

In general terms, dynamic range refers to the difference between the smallest and largest values of a quantity. In music, it means the difference between the quietest and loudest sounds. Audio compression effects operate on this dynamic range by reducing the difference; expanders do the opposite. Over-use of compression in digital audio has led to the ‘loudness war’. People who care about this stuff have developed various standards and measures of loudness and dynamic range, e.g. EBU, the K-System by Bob Katz, and many more. This particular plugin measures dynamic range (DR) by calculating the difference in decibels between peak and RMS loudness – the difference between the loudest sound in an audio file and its average loudness overall. In foobar2000, it writes this information to a text file (example below) and to audio file metadata, allowing you to sort music by dynamic range. This is for information only; it doesn’t change the sound at all, unlike ReplayGain.

ReplayGain is a system for dealing with variations in loudness in a music collection. It adjusts the playback volume of a track or album based on its RMS loudness. The RG value is the amount of adjustment in decibels relative to a reference loudness value. Whilst this doesn’t solve the problem of dynamic range, it does correct for the differences in perceived loudness. There is some correlation between RG and DR, though, because they’re both based on RMS loudness: tracks that are heavily compressed and very loud will have low RG and DR values, whereas tracks that are more dynamic will have higher values of each. Different kinds of music suffer from lack of dynamic range to different degrees, but in general a higher dynamic range is preferable. In general, DR 10 and above is OK, and less than 5 can be a problem.

So here’s a few selected results of scanning the dynamic range of around 400 albums. Most of them are in the mid range – about 300 of these are between DR 8 and 12. In the list I’ve put a few pairs of albums with the same DR but different styles: Talk Talk’s Laughing Stock – an album often cited as a good example of music with lots of space and dynamics – is equal to the classic rock of Back in Black at DR 13. Manitutshu by Mark Fell is equal DR to the sparse and quiet album Alina by Arvo Pärt. EVOL’s Hyperobject 2 has a surprisingly high DR of 17, given that it’s an hour-long relentless composition based on a single percussive sound (“flangey side stick“). But in that sense, it’s not too dissimilar to Drumming by Steve Reich, at DR 16. The simple drone music of ELEH has a very low DR, but the more complex combination of drones and textures by Hazard (BJ Nilsen) has a high DR. At the top of the list, SND’s delicate stdio album is just below a live recording of a symphony by Arvo Pärt. Symphony No. 4 is a single track that’s mostly very quiet except for a few louder sections, which is probably why it heads the list (high peak – low RMS = high DR). Ben Frost’s industrial music is perhaps over-compressed at DR 5, but Powell’s dance music sounds fine at DR 4. The lowest DR is heavily distorted music – Russell Haswell, Kevin Drumm, and music from Japan.

DR 22: Arvo Pärt – Symphony No. 4

DR 21: SND – stdio

DR 20: The Automatics Group – Summer Mix

DR 19: Ben Carey – Primrose [disquiet0005-layer]

DR 18: Chris Watson – Stepping Into The Dark

DR 17: EVOL – Hyperobject 2

DR 16: Mark Fell – Manitutshu
DR 16: Steve Reich – Drumming
DR 16: Arvo Pärt – Alina

DR 15: Fleetwood Mac – Rumours
DR 15: Hazard – Wood

DR 14: Buckingham Nicks – Buckingham Nicks

DR 13: Talk Talk – Laughing Stock
DR 13: AC/DC – Back in Black

DR 12: ZZ Top – Tres Hombres

DR 11: AFX – Analord 01
DR 11: Autechre – Amber

DR 10: Black Sabbath – Master of Reality
DR 10: Taylor Deupree – January

DR 9: Aphex Twin – Selected Ambient Works Volume II
DR 9: NoMeansNo – Wrong

DR 8: Autechre – Exai

DR 7: Napalm Death – Scum

DR 6: ELEH – Radiant Intervals

DR 5: Ben Frost – By The Throat

DR 4: Powell – Club Music

DR 3: Boris – Akuma No Uta

DR 2: Kevin Drumm – Sheer Hellish Miasma
DR 2: Russell Haswell & PAIN JERK – Electroacoustic Sludge Dither Transformation Smear Grind Decomposition nO!se File Exchange Mega Edit

DR 1: NHK – ‘Stomp_1’, from SND/NHK – Split

DR 0: Kevin Drumm – Purge

These DR values are calculated per album; it averages the dynamic range of the tracks. As a result, albums tend to be more centred in the list, whilst the more extreme DR values are likely to be single tracks or albums with fewer tracks. An album with more tracks has less chance of scoring very high or low because it would require every track to have a similarly extreme DR value. For example, the top and bottom albums have just one track each. But in contrast, SND’s stdio (DR 21) has 17 tracks, which makes it more of a rarity.

This isn’t a very scientific analysis, of course. This kind of measurement depends on how loudness is defined and which type of weighting is applied. Perceived loudness varies with frequency, and perception of frequency balance also varies with loudness (equal loudness contour). As albums get remixed and remastered, their dynamic range can change. File format can also affect DR measurements, and in this list is a mix of CD, MP3 and FLAC. If this analysis tells us anything, it’s that when it comes to sound quality, different levels of dynamic range are appropriate for different types of music. A wide dynamic range in music is a generally a good thing, yes, but having a low dynamic range doesn’t necessarily mean that it’s music done badly.

Here’s an online database of albums and DR values: http://dr.loudness-war.info/ It also has links to download tools for DR measurement.