I am duty bound to start a thread about the loudness war

[Ⓐaron]

I've been putting this off for quite some time, but it's something that pisses me off and deserves to be addressed. Modern music is just too fucking loud. This might sound like a strange complaint coming from someone who listens to metal, so let me explain.

An audio signal is made up of a waveform that bounces up and down between a specified range. The larger the range, the louder the signal. Digital audio is contained within a specific set of ranges - 16-bit audio, the standard for CDs, can faithfully represent up to 96 dB of dynamic range within its signal-to-noise ratio.

Somewhere along the line, audio engineers started to get the idea that louder was better. There is research of dubious quality (which has been refuted by recent studies) suggesting that records mastered to louder volumes sell more copies. This was probably a lot truer before radio stations and other music distribution services started applying their own signal processing that automatically normalises all music to the same volume. Nowadays, the only thing that happens is that the music gets processed twice, making it sound even more like shit.

Don't get me wrong; until a certain point, louder probably is better. A lot of early CDs didn't actually reach anywhere near the maximum amplitude reproducible by 16-bit audio and thus probably didn't represent the signal as faithfully as possible. However, you can only get louder a certain extent until you start hitting the floor and celing of what 16-bit audio can reproduce. And around 1995, this is exactly what started to happen.

The tipping point is generally regarded to be (What's the Story?) Morning Glory by Oasis. While an otherwise decent album, its production was marred by excessive cocaine use by all parties involved, including the engineer. Somewhere along the line it was decided that audio fidelity didn't really matter. Vlado Meller, who has gone on to ruin numerous other albums as well (the Red Hot Chili Peppers' Californication, Kanye West's My Beautiful Dark Twisted Fantasy, and several albums by Gorillaz being excellent examples) just mixed the album so loud that the tops and bottoms of the waveform were clipped off. For a visual illustration of what a clipped audio signal looks like:

It sounds exactly as bad as you would expect it to. The result is audible distortion on the sound, which affects the sound of the drums particularly negatively.

It just got worse from there. Morning Glory was mastered at about -8 dB RMS; Iggy Pop's 1997 remaster of Raw Power averaged -4. Until about 2008, mastering records as loudly as possible appeared to be an irreversible trend in the music industry. It even affected the Beatles; here's a .gif of four releases of "Something":

To be fair, not all releases are clipped. There are a number of more sophisticated technologies that simply force the waveform to adhere to a very narrow amplitude range instead of actually cutting the tops and bottoms off. Here is a comparison of clipping to limiting:

As time has passed, the use of clipping has been replaced by limiting to a certain extent, although there are some artists whose releases are still clipped to this date (Muse and the Mars Volta being two flagrant examples).

The trend got especially bad when it started affecting even bands like Genesis, the Doors, and Joy Division, who had been around for decades. Labels would start reissuing new masters of albums that would have no dynamic range at all.

What happened in 2008 to bring things to a head was the release of Metallica's Death Magnetic. This album was so atrociously clipped that it was completely unlistenable, and it prompted a media backlash. Even engineer Ted Jensen, who himself has been behind no small number of brickwalled mixes, criticised the mix, claiming that the audio was already brickwalled when it came to him. What compounded the media firestorm was the release of all ten tracks from the album on Guitar Hero 3, in which they sounded just fine, with no clipping or (unwanted) distortion whatsoever. This suggested that it was a conscious choice by the band (and Rick Rubin. Seriously, fuck Rick Rubin; he ruins every album on which he's allowed anywhere near the mastering board) to have the audio clipped. The audio was even horribly clipped on the vinyl. There is no small amount of irony to the fact that Metallica, the industry's anti-piracy poster boys, have released an album that can only be truly appreciated via piracy.

Morning Glory was one tipping point; Death Magnetic was another. The backlash against that album became too great to ignore. Metallica, predictably, didn't flinch in the face of criticism and refused to admit they released an inferior product, but the controversy over that album doubtless caused other artists to re-evaluate their positions. Devin Townsend, with his 2009 release Ki, remarked, "I officially pull my hat out of the loudness wars," and late in 2008 when Bob Ludwig offered Axl Rose three mixes of Chinese Democracy, Rose went with the one that had the least volume compression.

Some helpful resources if you want to learn more or help free yourself from this plague are TV Tropes, Wikipedia, Imperfect Sound Forever,Turn Me Up, and Loudness-war.info. The last of these is a database of dynamic range, in decibels, on a large number of commercially released albums. The higher the score, the more range.

Additionally, there are ways to edit audio to undo the effects of clipping to a certain extent. It is incredibly time-consuming, requires a decent computer, and requires you to know what you're doing, and it still won't sound as good as if the album hadn't been clipped, but it will still sound better than the commercial release. If there is interest I can write up a guide on how to do this in Audacity.

 

[Ghost X]

I thought this was gonna be a thread about how humans are fucking around with birds and whales' lives (etc), because of how loud we are, but... I was wrong =(.

I think the increasing loudness in music is compensation for the decreasing quality of music. I challenge anyone to name a modern day song that'll be played in 30 years. 99.99% are forgettable, made merely to make money.

 

[Glaurung]

You are right, Aaron. Today at the bus a kiddo was playing "music" with his cell so everyone had to endure the torture listen to it. The songs were supposed to be rap and reggaeton, I guess, but I could only heard a lot of saturated noise, with the same repetitive rhythm :/

 

[Ghost X]

Another thing I noticed about modern music, is that the beat to songs are a lot faster, and that percussion seems to be used to carry music these days rather than a decent melody. Take away the percussion in modern day music, and what do you have? Crappier crap.

 

[Ⓐaron]

Aaron's Guide to Repairing Clipped Audio with Audacity

This was requested at another board, so I might as well poast it here too.

The first thing you're going to want to do is make sure the album is actually clipped. The best way to do this is by looking at the waveform. If it looks like this:

Then it's probably clipped. You can zoom in further to see what the peaks look like. A clipped signal will somewhat resemble this:

Where you can see lots of signals hitting the peak right next to one another. (Audacity has a feature called "Show Clipping," but this will only work if the signal goes all the way up to 0 dB, and it's not always reliable).

Clip Fix will affect audio compressed with other compression filters, but it will not do so as much. If the audio used a distortion filter, you'll get quite a bit of recovery, though not as much as you would from actually clipped audio. If the audio used a peak limiter, you probably won't get any audible difference. Unfortunately the only way I'm aware of to tell for sure whether distortion or peak limiting was used is running clip fix and seeing how much dynamic range was recovered.

So you've determined you have clipped audio. You're going to want to make sure you have a good-quality version of the album before you start repairing it, because there is no point in repairing bad audio. Ideally you should be repairing from FLAC or some other lossless audio format, because applying lossy compression to audio already sourced from a lossy source results in egregious losses of quality that even non-audiophiles will be able to notice. Unfortunately, occasionally people upload tracks that are converted from lossy formats as FLAC, so you'll want to go to the spectrum analysis, which you can do by clicking the arrow on the top next to the "1.0" in the track view and selecting "Spectrum". Zoom out to 22k. If you have frequencies going all the way up:

Then it's probably legit. (Occasionally, albums are mixed from lossy sources. In these cases, there is no way to get a legitimate lossless rip, unless you happen to have the original master tapes and can make your own mix; in which case, it would be pointless to Clip Fix the audio anyway. Also, with piano music and other instruments that don't have lots of high frequencies you won't see audio extending to 22 kHz, but piano music is rarely clipped anyway). The best way to ensure you have a legit rip is to make your own FLAC rip with Exact Audio Copy. If you don't have the CD, there are plenty of places where you can find rips, but teaching you how to pirate music is beyond the scope of this guide.

Ideally, the best way to repair clipped audio is to do an entire album at once. This way you can be sure you are reducing the volume of all tracks by the same net amount, meaning that you will not get inconsistent levels from track to track. I would recommend loading the entire album into foobar2000, selecting "Convert", outputting as FLAC, destination "Generate multi-track files." This will embed a cuesheet into the file that will preserve the track listing, and it will also enable you to work on the entire album at once.

First I am going to run you through how to fix a case of ordinary digital clipping, and then I will run you through a couple of special cases.

The filter we're going to be using is called "Clip Fix." You can access this under the "Effect" menu. Clip Fix advises you to reduce the volume of the track by 10 dB. In my experience, this often isn't enough. If an album is as badly clipped as My Beautiful Dark Twisted Fantasy (the record I'm using for my example), you will easily recover much more than 10 dB of dynamic range. We're not going to be keeping all of this, because Clip Fix often overcompensates, but you need the room to work with in the mean time. I would recommend amplifying by -20 dB. If you've over-compensated, you can always re-amplify later. So the first thing you'll want to do is select the entire album, select "Effect - Amplify" and type in "-20." (You can use a different number, but if you do, you'll want to remember it and replace future references to "-20" in this tutorial with the number you use).

After you've got the track volume reduced, you're going to want to make sure the audio maintains a consistent level across the album. Sometimes record companies clip different tracks, or even different sections of the same track, at different levels. I have no idea why they do this, but I can name numerous albums by different artists (Dream Theater - Falling into Infinity, several albums by Moonsorrow, etc.), where different tracks, or different sections of the same tracks, are clipped at different ranges. You will want to zoom in at the peak of the waveform and see if they are clipped at different ranges. Here is Falling into Infinity by Dream Theater:

As you can see, the second song is clipped at a different level than the rest of the album. So for Falling into Infinity, you will want to start by clip fixing the first song, then clip fixing the second song, then clip fixing the rest of the album approximately twenty minutes at a time. The reason you need to do this is because Clip Fix only addresses audio near the peak of the selection. Since the clipped audio in the second track will mostly not be near the peak of the selection, you will need to address the audio in the second track separately.

The settings you're going to want to use with Clip Fix will usually be the default. (In cases where faded out audio is clipped, you may wish to adjust the settings; I will address doing faded out audio later). 95% threshold is good. Too much lower than that and you start having the filter alter audio that's not really clipped. That's bad. I would say 90% is the absolute minimum you will want for audio clipped at a constant level. In some cases you actually may need to go higher (98% or 99%) to preserve audio fidelity. I have not actually worked out why this is, or how to identify recordings where this will be necessary, but when I did a 95% clip fix of Petrychor's Dryad I noticed a distinctly audible loss of higher-frequency registers that did not occur when I clip fixed it at 99%. This is the only time I have ever observed this happening, however.

Once you've done the entire album, you will have something that looks like this:

Unfortunately, Clip Fix isn't a perfect filter; it will have overcompensated in many cases and created audio that will create undesirable "pops" in the speaker. If we've recovered more than about 5 dB of audio (you can tell how much you recovered by going to "Amplify" and seeing how much it suggests adding to the piece; subtract from 20, or whatever number you reduced the volume by, to see how much you recovered), we'll probably have this problem. The best way to deal with this is (strangely, given how much I usually hate dynamic range compression) by applying moderate volume compression to the repaired segments of the audio.

There are a number of different compression algorithms one can use, but I find most of them to be flawed in some way or another. There is a custom "soft knee" filter that does not work instantaneously and produces some distortion; there are several compressor filters that usually create audible "pumping" in the mix due to their lack of instantaneous attack and release. The best compression algorithm I have found in Audacity is the default plugin called the "hard limiter," though it is actually a hard knee filter.

Remember how we reduced the volume by -20? We're going to start at .5 above -20 and keep periodically applying the hard knee filter. Go to "Effects - Hard Limiter", input "-19.5" as the dB limit, and input something close to 1 as the "Residue Level." I select .840896 because it is the quad root of 0.5, meaning you will cut the amplitude in half approximately every 2 dB, but it is not necessary to use something so pedantic. As long as it is fairly close to 1, the effects will be desirable.

Now keep increasing the limit by .5 and repeating the filter until you hit the peak of the audio. You can tell what the peak is because it will be specified in the "Amplify" filter. If the value of your last pass through the "Hard Limiter" filter is greater than the value given to you in "Amplify," you're done.

Now the last step is to amplify the audio. I would recommend leaving 0.1 range, like so:

(If I had done the whole album we would be amplifying by much less than 14.2 dB; "Runaway" is one of the quieter tracks on MBDTF).

Naturally, since we sourced this from FLAC, we'll want to export this as FLAC. If you save as sixteen-bit FLAC, you will actually probably still be throwing away a small amount of audio data, since the volume has been reduced and the amount of data we have is beyond the scope of the sixteen-bit sample rate, but 24-bit FLAC is usually about twice the size of 16-bit FLAC and it is unlikely that the human ear can tell the difference with a clip fix from a 16-bit source, so I would recommend simply exporting as 16-bit FLAC.

That said, certain methods of exporting 16-bit audio result in the introduction of a fair bit of noise into the files. It is probably not audible to most people, but it will make the files larger. Before you export, I would recommend going to "Preferences", selecting "Quality" and setting "Dithering" to "Rectangle". This will eliminate the problem.

And we're done. The result should look something like this:

And it should be much more pleasing to the ear than the crap record companies usually excrete these days. Note that we will not have fixed all instances of clipping in the album; Clip Fix is not that good. But we still will have solved enough of them to make a difference.

Two additional last things are worth noting. The first is cases where engineers clipped the audio before fading it out. This is a massive pain in the arse and unfortunately, the way the plugin exists right now, the only way to deal with it is to clip fix a few tenths of a second of audio at a time. This example from Moonsorrow's Kivenkantaja is one such case of clipped fadeout:

And here's a zoomed-in view of one peak where you can tell the peaks are clipped:

Unfortunately, you will simply have to bear the tedium of applying the filter several dozen times. Luckily, since you're only clip fixing a few tenths of a second of audio at a time, you won't have to wait long for the filter to execute. You'll also want to make sure you're not selecting near the peaks of the audio, as Clip Fix will not affect audio near the edges of a selection. Make your selection between peaks. Also, with faded out audio it is fine to use a somewhat lower threshold of clipping, although you're not going to want to reduce it too much. Sometimes, if your sensitivity is too low, Audacity won't even recover a lot of clipping. I would recommend using 90% as the threshold for faded out clipping. (Just remember to increase it back to 95% when you go get around to fixing the next track).

The only other thing worth noting is cases where there is "soft" or analogue clipping. The remasters of Van der Graaf Generator are like this. You will see waveforms approximately like this:

Ideally you're going to want to go back to non-clipped sources (like the original VdGG CD issues) because this is a massive pain in the arse to deal with, but otherwise the only real way to deal with it is by clipping the audio at a constant level. You can do this by using the "Hard Limiter" as an actual hard limiter, with a residue level of 0. You can tell what decibel level you want by selecting "Waveform - dB" in the audio view dropdown menu and zooming in. Ideally you'll want the smallest possible value that results in clipping at a constant level. Pawn Hearts complicates things further by clipping each track in a different manner (some are clipped at a constant level and some are clipped in a more analogue-like manner), so this is one of the rare cases where you will want to handle each track separately. Just make sure you amplify each track by the same amount, or albums that are continuous will have undesirable jumps in volume at track divisions.

You can create your own lossy encodes of the music after you've exported it if you want to put it on an iPod or something, but I'd recommend saving the FLAC (burning it off to disc if you're running low on disk space) so you don't have to repeat the process, because it is an incredibly time-consuming one.

Let me know if you need clarification on any of this and I will try to do so as best as I can.

 

[Cat Rage Room]

I challenge anyone to name a modern day song that'll be played in 30 years. 99.99% are forgettable, made merely to make money.

To be fair; a lot of them. Kanye West, Jay-Z, Alicia Keys, Beyonce, etc etc. Mainstream or no, they're pretty fucking notable and a lot of modern musicians break ground. Remember, Michael Jackson, the Beatles, Tupac, and Elvis were mainstream once and they're the most acclaimed and remembered modern musicians who ever lived.

You're looking at the the wrong way; the obscure stuff, no matter how good it is, that's the stuff that won't be remembered. The mainstream stuff is the stuff that survives history; unless you want to believe that in 30 years there will just be a a blank amnesiac gap in musical history or something?

 

[Ⓐaron]

There's definitely lots of great music being created right now, including the stuff Mog mentioned. Unfortunately you often have to dig beneath the surface of what gets played on the radio to find it. Even with stuff like Kanye, often times his best cuts don't get played on the radio. In my opinion the two best cuts off his latest solo album are "Blame Game" and the extended version of "Runaway," and neither of them is likely to be played in its entirety on the radio.

Hip-hop is almost a unique case in that a lot of good stuff right now is still commercially successful (although by no means all of what is commercially successful is good). With most other genres, you don't have that anymore. Most of the rock and metal that gets played on the radio is shit. You have to look to stuff issued on indie labels or less promoted by the majors to find quality anymore. And, of course, classical and jazz have completely disappeared beneath the radar, despite the fact that there is still plenty of innovative material being released in both genres.

 

[Ⓐaron]

by the way, a couple of audio illustrations:

 

[Cat Rage Room]

There's definitely lots of great music being created right now, including the stuff Mog mentioned. Unfortunately you often have to dig beneath the surface of what gets played on the radio to find it. Even with stuff like Kanye, often times his best cuts don't get played on the radio. In my opinion the two best cuts off his latest solo album are "Blame Game" and the extended version of "Runaway," and neither of them is likely to be played in its entirety on the radio.

Hip-hop is almost a unique case in that a lot of good stuff right now is still commercially successful (although by no means all of what is commercially successful is good). With most other genres, you don't have that anymore. Most of the rock and metal that gets played on the radio is shit. You have to look to stuff issued on indie labels or less promoted by the majors to find quality anymore. And, of course, classical and jazz have completely disappeared beneath the radar, despite the fact that there is still plenty of innovative material being released in both genres.

Excellent post, dude.

The above statement/question of 'Nobody will care about popular musicians in the future' always irks me (even if I understand the sentiment) because it basically boils down to 'WHAT WILL BE REMEMBERED: THE STUFF EVERYONE KNOWS OR THE STUFF NOBODY KNOWS'

I mean come on homie.

 

[Ⓐaron]

I also feel compelled to point out that music has been "going downhill" forever. In the fifties and the sixties it was those kids and their rock music, in the thirties it was those coloureds and their jazz. In all cases there has been plenty of music that has endured and attained classic status. True, there's a lot of stuff played on the radio now that no one will care about in ten years, but it's always been that way. Almost no one listens to the Dave Clarke Five now, but at one point they were as popular as the Beatles. Sure, there's lots of popular stuff now that won't endure, but it's always been that way. Music will survive, no matter how much the industry tries to ruin it.

 

[Cthulhu]

I for one believe we'll be listening to the same crap in thirty years time . Pop music from the 70's, 80's and 90's is still on the radio regularly - with whole channels dedicated to them - and they make regular re-appearances in movies and TV shows too. I'm pretty sure the same will happen with mainstream music of this age.

And I don't mind if what we consider to be better music remains somewhat underground. That way, it won't be popular, and there will be less kids fucking it up.

 

[Hisako]

I always had a feeling todays songs were mixed fairly loudly, but now I look at those Audacity screenshots?

Holy shit that is ridiculous. In another era that would just be considered a metric fuckton of noise.

 

[looneymoon]

As far as I'm concerned Arcade Fire's "Wake Up" is THE song of the generation and I have yet to see anything to change my mind. I'll be honestly surprised if no one remembers that song 30 years down the road.

Also I'd like to point out that there are "mainstream" popular bands that are still acclaimed in indie markets as well. I mentioned Arcade Fire, but take a look at bands/artists like Kanye (as mentioned before me), RADIOHEAD (if anything else, people are gonna remember Radiohead), The Gorillaz, Florence and the Machine, Outkast, Janelle Monae, Amy Winehouse, The Strokes etc.

Popular=/=bad. New=/=bad, either.

Personally, if I discover a band/artist that isn't that popular it makes me really upset because it means less people I'm sharing it with. Then I wish that they were popular and more people could understand how amazing they are I don't really like this notion of underground = better because I think music is one of the most basic communal things that links people together. If I feel particularly lonely or antisocial, it's easy to just put on a song and it feels like someone's there - because someone else has heard that song and felt something about it too.

 

[Ghost X]

Okay, I concede my point. Sounds silly now that I read it again. There are good artists these days who produce timeless music, but there is a lot of shit out there on the airwaves. I think that was what I was trying to say .

As for the topic, there was a time when an old flatmate of mine would tell me how he hated clipped music, and he'd only download extremely high-definition music of his favourite artists, and mock me for listening to 'poor' quality stuff

 

[Ⓐaron]

I am not sure anyone cares, but I introduced a paragraph into my guide for clip fixing audio about the settings one should use for exporting 16-bit FLAC. It reads:

That said, certain methods of exporting 16-bit audio result in the introduction of a fair bit of noise into the files. It is probably not audible to most people, but it will make the files larger. Before you export, I would recommend going to "Preferences", selecting "Quality" and setting "Dithering" to "Rectangle". This will eliminate the problem.