Volume Wars: The Degradation of CD Mastering
by Eric Portelance
CDs were supposed to be a
complete revolution in audio delivery systems. In a sense, they
were. The shift to a digital medium was a great leap for audio
fidelity, and the slow conversion to digital recording and mastering in
the years that followed also helped. The quality of CD audio
peaked in the 80s, however, and it's been going completely downhill
January 13 2006
Here's the thing: Record label executives have been fighting a type of war for years now. Let's call this a volume war. There is this ridiculous idea that the louder you master an album, the more people will enjoy it, or the more it will be noticed. Now, there are many problems with this. To begin with, CDs have a very large dynamic range, meaning the level of individual frequencies that it can reach. This allows for more natural and precise replication of sound. However, when you master a CD at an ever-increasing volume level, a phenomena called "clipping" occurs. In simple terms, clipping is when the peaks (or a certain frequency) of an audio wave is cut off because it is out of range.
Clipping and Distortion
If you understand what digital is (binary digits), then clipping can be represented as the binary maximum that the system can support (all 1s!). However, the problem with this is that audio levels should never reach the binary maximum. If it does, it means that the frequency peak should technically have gone higher than this, but it has been cut off or limited. Now, by increasing volume during the mastering process, you are increasing the number of clipped peaks, and therefore reducing the quality and detail of the audio recording. It results in a far less natural sound.
There are two side-effects of this. First of all, the majority of us use transistor amplifiers, which are much like binary (either open or closed). The major problem with these types of systems is that they clip very harshly. Transistor amplifiers will produce a harsh distortion of the audio signal. This can sound somewhat static-like, or there can be other types of audio artifacts introduced which are not part of the intended audio signal. The more of this that you have, the less natural the audio sounds, and more detail is lost.
This is a part of the second side-effect of distortion. Have you bought or listened to a recent CD (especially in the past 3-4 years) and found that the sound was "tiring"? It's something almost subconscious, and difficult to put your finger on, but the audio appears to be very harsh, and something is off about it. This is the prime symptom of a poorly mastered album. The sad thing is that more and more albums today are suffering from this, and the hard work put into the recording and mixing processes are completely destroyed.
All this had led to an increasing degradation of CD quality through the years. Much as vinyl suffered from a problem of poor materials and mastering (to make them cheaper), CDs are suffering the same fate in the digital area. Vinyls that were mass produced tended to use a large percentage of recycled, and softer vinyl. This would lead to warping, and a quicker degradation from the needle.
In the early days of vinyl, though, and even toward the end, there tended to be great pride in the mastering process. Masters were handmade, and those who did the work often carved their initials or other personal symbols into the discs in order to mark their work. This sense of pride is no longer evident in the world of digital. Clearly, those who master albums know exactly what this is doing, but if they refuse to do what the executives tell them to do, they won't be working for much longer.
ReplayGain to the rescue
The fact is, there is absolutely no reason to master CDs this loudly. The entire purpose of an audio amplifier is to boost the sound to levels which you can enjoy. In doing this on the CD itself, you are in fact making it sound much worse. The rule of thumb is that over an average gain level of 89db, clipping will most likely occur.
What is gain? In the simplest explanation, it is the level at which a signal is boosted or recorded. In a slightly more mathematical explanation, gain is the ratio of output to input in terms of signal power.
There is an emerging standard called ReplayGain. Essentially, this will analyze a song or album and determine the relative level of gain through quantitative analyzation of the audio signal. A program called MP3Gain is an implementation of ReplayGain, and does this quite well. It analyzes your songs/albums and losslessly overwrites them with an adjusted gain level. This is important for two reasons.
1) By adjusting all your songs and albums to a same relative loudness level, you will no longer have to reach for the volume control every 30 seconds when listening to music of varied genres. Obviously, though, a flute solo and a drum solo are not meant to be played at the same level, and ReplayGain does take account of this (and the original intentions of the artist).
2) By adjusting gain to a level that allows a far greater dynamic range for the music, you are eliminating all sorts of amplifier clipping. There are separate problems of mastering compression and unintended distortion actually in the recording, which ReplayGain cannot fix, but your music will sound far more natural than it did before. Drums are especially noticeable, as those tend to be what produce clipping most frequently.
In looking at my CD collection with MP3Gain, I noticed that the average album gain is about -7db, meaning that it is 7db above the 89db standard gain level. Even a 1db change in terms of psychoacoustics can make a big audible difference. The worst culprits that I found were the horribly mastered Red Hot Chili Peppers - Californication, which features songs recorded at ridiculous levels well over 110db, and the new System of a Down albums (though they are all pretty poorly recorded, the newest ones have album gains of almost -20db!). This is completely unacceptable. Just take a listen to the song Parallel Universe on Californication (especially the chorus) in order to see just how bad this can get. In comparison, one of my favourite albums, Supertramp's Crime of the Century, is mastered at a dead-on 89db, and so it has an album gain of 0db. I have the original CD release of this from the 1980s, but I am sure that subsequent remasters have boosted the gain and introduced clipping, as is the case for almost all remasters of classic albums.
The following two very unscientific and unnecessarily small screenshots of Winamp's spectrum analyser show the same song before and after ReplayGain was applied. These are both of exactly the same timeframe in the song, so you can see the much wider dynamic range afforded and the detail that is restored in the signal. For those who don't know, when the bars hit the top, that indicates clipping. In the "after" shot, the bar that hits the top is not actually clipping, but rather it is maxing out the available range without going over it and clipping.
The Importance of Album Gain
MP3Gain can either adjust an individual song to 89db, or it can look at a whole album and bring the average down to 89db (meaning that some songs will likely be above or below this level). Album gain is an extremely important concept because, on albums, some songs are intentionally mastered louder or softer than others, for emotional effect. This is further important in the case of albums like Pink Floyd's Dark Side of the Moon, on which all the tracks segue (or blend) into each other as a seamless whole. If all the songs were adjusted to the same volume level, there would be noticeable differences when the tracks would change, and this would alter the intent of the artists and distract from the music.
Levelling vs. Peak Analysis
As I have already explained, the ReplayGain standard uses algorithms that analyze the psychoacoustics of a specific song, as opposed to simply calculating an average volume level. This is important because the way in which the brain perceives sound is not necessarily identical to the theoretical concept of it. Peak Analysis also does not fundamentally change the intended volume differences in a song.
Most programs nowadays can perform a type of track-based volume levelling. This essentially calculates an average and sets all tracks to this average. The major problem with this is that changes in volume are often intended in a song, and this cancels out that effect. Volume levelling is the standard that is used on the radio. Each song should be played at exactly the same level, whether it is Vivaldi or Slayer. A slow piano solo in a song should be played at a similar level as the distorted guitar solo. Levelling also does not take account for specific mastering differences amongst tracks in an album, as I explained before.
Theories and Conclusions
Why is this happening? My theory on the matter is related directly to the flood of cheap electronics on the market. People don't give a rat's ass about good sound, and they don't even know it when they hear it. As I said before, amplifiers are meant to boost volume to your level, and so it should not be ruined on the medium. However, in the case of cheap electronics, boosting the volume level on the recording is advantageous because these electronics can probably not pump out sound that is loud enough for most people. There would obviously be complaints that the volume is too low. This also created a double-edged sword. Cheap electronics often distort at high volumes to begin with, and so boosting the recording level on these makes the sound even worse.
The music is perceived as "better" by those people who have music systems incapable of producing the full frequency range at a relatively even sound pressure level. Portable music players like the iPod are also guilty. The same people who think the included 50¢ earbuds are "great" are also the ones who love to play their music as loud as they possibly can. Being a portable device, the iPod can't produce the loudest volume out there, in order to conserve battery power and for other reasons that I won't get in to. Seemingly, a solution to this is to boost the gain level during mastering, so that the amp is not working any harder, but people get that loud sound that they want, and they can even push it higher.
There doesn't appear to be an end in sight to this madness, and I suppose the record companies think they are just giving people what they want. I believe, however, that if people were more sensitized to this issue, that there might be an urge to bring that volume back down again. It's only a matter or releasing more and more poorly mastered albums before people start wondering why they sound so horrible. As an example of this, my mom bought the new (horrible) INXS album. To my dad's awe, the bass is so ridiculously overblown and distorted that it is almost unlistenable at any length. Out of curiosity, I analyzed the gain values of the album and, sure enough, a huge percentage of the frequencies are completely clipped off. The frequencies around 60hz, in particular, are just clipping non-stop. As sad as it may seem, it will take more of this crap to get people to wake up so that we can enjoy good audio again! On that note, I'm going to pop in Crime of the Century in order to really go back to what great mastering sounds like.