The Loudness Wars and Dynamic Range
Date: May 20th, 2013 @ 12:14
The Loudness Wars is a struggle between the desire to capture a high fidelity audio source in a commercial recording, and the push—mostly by marketers—to increase the overall volume in what is often regarded as a way to trick consumers into thinking a given pop album will sound better to most customers.
Increasing the volume of music often sounds like a much better idea than it is, as it comes with a cost. Every distribution format has its own drawbacks and benefits, and when music is distributed at increased volume, the trade-off is that in order to accomplish this increase, the dynamic range has to be reduced. Ultimately, people control their own volume knob, and you can’t reliably “trick” them into playing something louder. In the Triassic period, when the dinosaurs were first getting their bearings and people still watched television according to network programming—and there were only three, possibly with NET/PBS comprising a debatable argument for the existence of a fourth network—television programs were interrupted on a regular schedule by video advertisements or “commercials” (unfortunately, YouTube has recently re-discovered this technology). It was not uncommon for commercials to be programmed louder than the programs they interrupted, presumably because since this species of early mankind hadn’t developed TiVo or Netflix yet, there wasn’t a pause button; one had to save up menial tasks like bathroom breaks and snack preparation and swiftly accomplish these goals during commercial breaks. Advertisers needed to cut through your distraction by shouting you down in your own home. By the same logic, it was useful to try to get your song played louder on the radio (which was often tricky, since radio broadcasters generally weren’t enthused by such schemes), or to make their album more appealing for car play by having a louder presence.
The comparison NPR makes to graphics in a 2009 article is an interesting analogy, whereby the increase in volume and decrease in dynamic range is turned into a visual metaphor about pixel size. Unfortunately, it misses the mark a bit in that it equates a decrease in dynamic range as an increase in pixel size, which can be confused with enlarging a photo; this is something you want to do from time to time. I believe a better analogy would be to equate dynamic range, which affects the contrast of the loudest portions of a song to the average and quiet parts by effectively lowering the contrast and increasing the overall volume, to visual contrast.
For the purposes of this analogy, please consider “loudness” as equivalent to “brightness”. Presume that you’re a “savvy” record label executive, recording engineer, or even an artist and you think this compressed dynamic range is the best thing since sliced bread, since it makes your pop music louder than anyone else’s pop music. This impetus comes from all end of the record label food chain, of course, and even artists demand it if they’re not aware of the pitfalls. Some of those who are aware of the costs still demand it. More than anyone else, it used to be the job of the audio engineers to know better, but it’s a competitive world and eventually they’ll come around to your way of thinking if you threaten to start taking your business elsewhere. You’ve got everyone on board: louder is better! Brighter is better!
Let’s pretend that this modest little photograph of a piano keyboard represents your artistic masterpiece. The top row represents your art as performed.

I’m going to mix audio and image terminology fairly often here, since the latter is acting as a metaphor for the former. You want your album louder since your marketing department tells you that sells more units, so by raising the brightness by 150 you get a “louder” sound. These numbers come from Photoshop adjustment filter level settings. The problem is, most of your customers aren’t going to listen to your album louder, they’re going to adjust the volume to whatever they like, which will result in a sound with less dynamic range played at any given volume. This is represented by the “+150 -150″ square in the second column. If you look closely, you’ll notice that the peak levels (the whitest whites) “top out”, making the brightest highlights on the image a bit flatter. The “difference” column represents the altered image in the second column (or the audio as played) from the original photo (or the audio as recorded with a decent dynamic range). Depending on your monitor settings, it might be subtle, or it might be immediately noticeable. This is relevant, too, since the same can be said for audio - some people will be listening to the album on an MP3 player at 128 kilobits per second on an iPhone with stock earbuds, and some people will be listening on a home stereo or PC with better output, and listening to your album at a higher bitrate or perhaps even using a lossless codec if they’re not using the physical media.
The catch, however, is that your competitors aren’t going to let you have the magic of artificial loudness to yourself, so this escalates the Loudness War. If a little is good, a lot is better, so you convince your audio engineers that your product Needs More Loudness! This brings you up to a brightness increase of 300. When your customer plays it back at normal volume (second column), the output is flat, and it’s really lost a lot of character at the high end and the low end is suffering, too. It’s certainly louder, however! It’s also a lot different (third column), having been stripped of most of its subtlety in the process. It’s not as bad as it sounds, however, since this is also the case for a lot of your competition’s most popular output, but now everyone’s on the same level.
You have a brilliant idea how to counter this, however. You just turn the volume up to eleven, and what you’re left with is the loudest, brightest album yet at +450. When your customers play the music, it is devoid of subtlety, everything is loud, and if they complain, that’s fine because they just don’t understand, man! By now, the album has had almost all of the dynamic range removed from it; drums sound like metal garbage can covers, guitars sound like they’ve been put through a series of compressor pedals, the vocals are always shouting, and your album will out-blare anyone else.
At this point, you may think that at +450, this whole brightness contrast and audio dynamic range metaphor is broken and that I’m taking the argument to absurd levels. In a way, I am. But so are some of the record companies, audio engineers, and even artists. Metallica released Death Magnetic in 2008, and there was a bit of discussion about how flat the album sounded. These claims were largely ignored. Without a source with more dynamic range to compare it to, it’s fairly difficult for most normal music fans to express what’s “missing” or what’s “different” about a given album, especially when all the obvious parts are there such as Lars’ fast drumming, Hetfields growly vocals, Hammett’s guitar solos… oh wait. And Jason Newsted’s bass was… oh come on, too soon for that joke, too? Snarkiness aside, the album was well-reviewed and considered by many fans either a return to form, or the continuation of a return to form first suggested by St. Anger. My point isn’t to tell you to love this album; I gave up on Metallica shortly after 1991, along with most traditional and thrash metal. What I’m more interested in getting across is that for people who considered themselves Metallica fans at the time Death Magnetic was released, this was a very well-received album. The fans generally embraced the album. Then the album was released as downloadable content for Guitar Hero III: Legends of Rock, and many fans had a more concrete explanation for what they were missing: dynamic range. An image on the Wikipedia page for the album shows the compressed waveform which has had its dynamic range reduced to the point where the waveform looks flat or “brickwalled”.
The colourful heat map chart above needs some additional explanation. The highest percentage of albums from a given year as examined and reported on dr.loudness-war.info are represented for each year. The data is a bit messy, but overall, there’s enough data to get a lot of interesting information out of it. The first column is the sum of all albums. The second column represents albums with unknown dates, and there are so few entries before 1970 that it makes the most sense to group the 1950s and 1960s as one decade column. Given how little data there is until about 1984 and that most of it is represented by later CD-era reissues not correctly labeled by reissue date, take the older columns with a grain of salt. In the last several years, there’s been a lot of poorly-reported information regarding media types and sales. The story a lot of people have been reading over and over is how the CD is dead and how vinyl is on a comeback, but it’s important to take these reports in context. Vinyl sales are up since 2007, and way up since 2011, showing no signs of waning by mid-2013. These stories aren’t wrong, however they almost always fail to mention that when compared to CD sales, these sales figures are roughly equivalent to 2-3% of CD sales in an era where CDs are doing quite poorly compared to their sales peak period. I say this to point out that the media type availability at the top of the graph are there to point out when various music distribution types have been mainstream and popular, not when they’ve been available. Similarly, although the MP3 was officially invented in 1993, AAC in 1997, and FLAC was released to the general public in 2001, digital distribution didn’t become mainstream and popular and relevant to sales figures until Apple created a brand and customer base that was important enough that record labels could no longer ignore the issue, or could no longer putter around with a web store while they payed catch-up. So, the end date of any particular format on the chart is meant to reflect meaningful sales, not merely availability or poorly-researched news articles.
One would hope, naïvely perhaps, that a word like “remaster” means that an album has been remastered with the intent at providing a high degree of dynamic range, but this isn’t the case. In other words, it’s not only new music which suffers, but reissues by people who should know better. The DR values generate a number based on an algorithm only an audio engineer could love. Higher numbers represent a higher degree of dynamic range. A lot of misinformation is put forth in many forums about how CDs have awful dynamic range, how MP3s have awful dynamic range, or that vinyl has greater dynamic range. This is wrong in as far as capability is concerned. It’s possible to get a reasonably high amount of dynamic range on a cassette MP3, when ripped and encoded properly. CDs and vinyl can both have fantastic dynamic range, but the loudness war has affected the CD more not because of technological limitations, but because of use cases: people tend to buy less disposable pop music that they’re only casually interested in on vinyl, and people who enjoy playing music in their c ar are limited to CDs and digitally encoded files - usually only MP3s unless one has purchased a particularly feature-rich after-market car stereo. In addition, there’s some self-selection bias in the data itself. Compact cassettes, 8-track tapes, and vinyl are notoriously laborious to rip compared to a CD, which is very nearly fire-and-forget, and even that’s a lot more effort than downloading, in many cases. These more difficult rips tend not to be random albums but the best of the best, the worst of the worst, and those albums where people can shop around for different versions of a popular release, and are curious enough to research the data on dr.loudness-war.info.
Several people and organizations actively attempting to lobby record labels about the issue suggest an ideal goal of having a Dynamic Range (DR) number of 14. The average of all the data present on dr.loudness-war.info (nearly 40,000 entries) is 9.87, and the median is 10. The medians are always within one point of the average, so unusual outliers don’t impact the data noticeably. It would be helpful to contextualize some popular releases, but the best approach is to examine the data source for albums a given reader is personally familiar with. Nevertheless, a few common points of reference are useful:
Iggy and the Stooges - Raw Power (1997 remaster): DR 1
Metallica - Death Magnetic (CD release, 2008): DR 3
Red Hot Chili Peppers - Californication (1999): DR 4
Pearl Jam - Ten (2009 remaster): DR 6
Michael Jackson - Thriller (2008 CD): DR 8
Daft Punk - Random Access Memories (qobuz “Studio Masters Edition” and Columbia CD): DR 8
The Beatles - Stereo Box Set (2009): DR 9 to DR 10
My Bloody Valentine - Loveless (1991): DR 10
Pearl Jam - Ten (1991 CD): DR 10
The Beatles - 1960s vinyl albums: DR 10 to DR 13
Iggy and the Stooges - Raw Power (1973 vinyl): DR 11
My Bloody Valentine - m b v (2013): DR 11
Beethoven / Karajan / Berliner Philharmoniker - The 9th Symphony (2007): DR 12 to DR 13
Metallica - Death Magnetic (Guitar Hero III DLC): DR 13
Daft Punk - Random Access Memories (Columbia vinyl): DR 13
Miles Davis - Kind of Blue (1959, various reissues): DR 13 to DR 15
Michael Jackson - Thriller (1982 vinyl): DR 15
Metallica - The $5.98 E.P. - Garage Days Re-revisited (1987 vinyl): DR 15
Metallica - The $9.98 CD - Garage Days Re-revisited (1987 CD): DR 16
Pink Floyd - Dark Side of the Moon: This album has been released countless times in various formats. Generally, official releases range from DR 9 to DR 12, with the DR 12 being one of the more highly regarded Japanese vinyl pressings.
Noise: Several noise groups are notorious for using a lot of highly-compressed noise, which sometimes results in very low DR ratings, such as Venetian Snares, Merzbow, Muslimgauze, Sunn O))), and Boris. Interestingly, Muslimgauze also has a release at DR 19 and DR 20, which is not only astonishing compared to other entries in the database, but uncharacteristic for his rather large discography.
Classical music: Classical music by esteemed symphonies and released on prestige record labels is often recorded quite well, and regularly scores from DR 12 to DR 15.
Several myths can be addressed with the Metallica entries alone: The Garage Days Re-revisited CD is a wonderful example of the CD media performing quite well in terms of dynamic range even when compared to vinyl when the two sources use the same master. There is nothing inherently wrong with CDs in this regard, nor many other regards. Clearly, vinyl can perform well, too. A more important point can be made with Death Magnetic, perhaps: artists are not always the best judge of sound. This statement will seem absolutely obvious to some, and counter-intuitive to others. Metallica has been playing loud music live throughout their entire career, and hearing damage comes with the lifestyle. When Death Magnetic was reported to have a sound that lacked dynamic range, their reaction can be summed up quickly: it sounds like it’s supposed to sound. When the Guitar Hero III: Legends of Rock DLC of Death Magnetic was released, Metallica initially underplayed it, then finally decided it didn’t sound all that bad anyway and besides, that’s what we pay Rick Rubin for, so we’re backing him up on this. The point isn’t that Metallica is dumb, or that metal bands play a type of music that doesn’t need dynamic range; the point is that artists and even name-brand audio engineers who ought to know better than compressing the hell out of an album often have no idea what they’re talking about. More charitably, they often don’t know everything they’re talking about, even when they damn well should.
Clearly, reissues and remasters aren’t always better, but they can be. Given that record companies are in the twin bind of both wanting us to pay for our favourite albums for a fift (or more) time while at the same time dealing with the fact that their business model is failing fast, they owe it to their customers to make remasters sound better than the original, or at the very least, not sound worse: flat, dull, and lifeless. It’s pretty bad time to be subtracting value in the music industry. Consumers had to treat the advantages of widescreen video as an education issue, and the Loudness Wars is no different. In the case of poor-sounding remasters in particular, when you spend your money on them, let the seller know if you’re not happy and why; don’t assume they’re familiar with the issue or that other customers will have already expressed their dissatisfaction.
More Data
I’d enjoy being able to do more with the data, but as it is, it’s not organized to make any definitive statements about media type, genre, or record label. Most additional analysis is based on taking the information users volunteer and put into the “album” field of the Dynamic Range database, and there’s often not much media specific data, but what there is can be isolated:
Average DR:
All entries (39,369): ADR 9.87
CD (2,980): ADR 9.80
Vinyl (2,290): ADR 11.89
SACD (288): ADR 11.77
DVD (249): ADR 10.40
Cassette (26): ADR 9.38 (if this seems high, consider the fact that most people who still rip cassettes are doing so with better equipment than most people had in the era when they were still popular)
Original (389): ADR: 10.77
Master (840): ADR 9.55 (attempt to generalize remasters, as original masters are less likely to be specifically referred to as a master)
(i) (25,908): ADR 10.00 (this represents those entries with more robust information - users took more time with these, generally)
lossless (34,372): ADR 10.00
lossy (4,997): ADR 9.01 (I can’t put too much stock in this. Users who keep ripping logs are often doing so in conjunction with using Exact Audio Copy for lossless rips, and less prone to be posting old, poorly encoded lossy files. In my own tests, I’ve sometimes found MP3s to vary from lossless by ± 1.)


Google Instant is that Google feature which suggests results as you type. For example, should you type “google is” and wait a fraction of a second, you’ll probably see something very much like the image at the right. ”Google is god” led me to wonder what else religions might be. I’m not suggesting this is in any way 



