There has been a lot of talk about dynamic compression for far too many years. Dozens of sites, articles and forum discussions have covered the subject time and again. Several studies have found no evidence that compressed music sells better, yet producers still persist in pumping up their recordings. The majority of music listeners, when given the choice, prefer the sound of dynamic music, yet producers still persist in pumping up their recordings. Tracks played on radio, air or internet, are normalized to the same level for broadcast, yet producers still persist in pumping up their recordings. Why does it continue? I can't explain why the producers, who still persist in spite of the facts, are such spineless monkeys. I can explain with a vivid illustrated analogy, what it is, what it means, and why it slips under the radar.
What it is. Dynamics are the range and rate of change from soft to loud (volume level) in music or an electrical signal. Dynamic compression is the amplification of the average level of an audio signal. In order to increase the average level and not go over the 0 dB mark (maximum allowable level), a limiter kicks in to keep the signal at or below 0 dB. In the process, the dynamic range is reduced—the more amplification, the more compression. Transients get hit the hardest because they are very strong, short, sharp peaks on the attack of a note, such as, the strike of a drum, the pluck of a string, the crash of a cymbal. High compression ratios narrow the dynamic range to the point that nothing is left but loud and louder.
What it means. Dynamics and transients are an integral part of music—as essential to musical expression as rhythm, melody and harmony. Taking away transients and reducing dynamic contrast diminishes the power of music.
Why it often goes unnoticed. Voices and instruments have a different quality of sound when played soft or loud. Our perception recognizes loud even when it's played back significantly softer than the original performance. The reverse is also true. The sound of a whisper is distinctive despite the playback volume.
Dynamic compression used to be applied to every recording. Analog recordings required it. Today, with the wide dynamic range capability of digital audio, it rarely needs to be done for most music, yet there are some exceptions. If done for the right reason, judiciously applied, and only when necessary, it can be a useful tool for making high quality recordings. Applied on the playback end it has practical applications for listening to music in noisy environments, as in cars, airplanes, busy urban areas, or for background music. Keeping levels in an artificially narrow range under those conditions makes for better listening. Far too often, though, modern recordings are being sledgehammered with dynamic compression gone gonzo. The sound quality is compromised. The music is compromised. And it's a sneaky little devil, because as noted above, it isn't always immediately apparent or sound terribly bad.
The music you buy ought to come complete. It should be delivered with all the energy musicians infuse into their music, fully intact and uninhibited, with all the expressive punch, with every milliamp of contrast and power. If later you choose to reduce the dynamics for portable listening, that's fine. You should be in control of when and how much. Most digital music players have a setting that adjusts tracks to play back at relatively the same level. The problem is most recordings of popular music come pre-compressed. The choice has been taken away from you. In the final mastering audio engineers, who really do know better, are stripping away some of the music. They've decided that you don't deserve it all. You only deserve a flattened out, squashed down, distorted version. It's cheating the music; it's cheating your ears.
Let's look at what happens with a visual example. The image at right is a full scale image, that is, it has areas of the blackest black, the whitest white, and a wide range of smoothly transitioning shades in between. There is only a relatively small amount of black and a smattering of white spectral highlights. Detail is seen in the shades of gray; absolute black and white define the limits. Below is a histogram of the image showing the distribution of values from 0 (black) to 255 (white). It's dominated by grays; the black and white trailing off.
We'll start by compressing the highlights. The white and light shades get pushed down, in other words, limited. The darker tones also get compacted accordingly. The whole image becomes dark. The values in the histogram are now bunched into the lower half. There is a direct analogy between light and sound. Black is silence, white is the peak level, or 0 dB. Spectral highlights, the sparkle in eyes or reflections off surfaces, are the equivalent to transients. Take away the spectral highlights and the light grays, we're left with mud.
Now let's do what audio engineers do to music by pushing up the levels. The white is back, but it's just a uniform wash of white. The black becomes middle gray, and the spectral highlights in the water are now sparkle-less without darker tones for contrast. Everything is bright, but the limited contrast range makes the image dull. This shows you the visual equivalent of audio compression, and what "The Loudness War" has done to music; left it bleached out and flat. There are no winners in this war. Note how pushing all the tones up to high levels has increased the overall, average brightness, likewise done with the audio signal during compression. Now, you may ask, can the dynamics be restored? Well, we can try. However, a lot of information has been lost in the compression process. After previously differentiated values have been reduced to the same value, there's no way to separate out those lost levels or to recover the lost shading.
Here's an attempt to stretch out what's left after compression. The black is back, and although the highlights are no longer maxed out, neither have recovered their original detail. It looks weird, as if the image has been both over and under exposed. And look at the histogram. See the gaps formed by stretching out the compressed values. There are no more smooth transitions between shades. The texture in the building's stone facade is lost permanently. The highlights on the statue are shapeless patches. The surface of the horses has taken on a grainy look compared to the original. —
The same sort of thing happens to compressed music. Unfortunately, it's less obvious to the ear. And because it doesn't sound as obviously wrong as it looks visually, recording engineers get away with it. The following histograms are from three different music CDs.
The first pair of tracks is a typical example of compression. Note how the peaks get flat topped—and this isn't the most egregious example.
Second pair is a properly adjusted master. Peaks are hitting the maximum, but nothing is blocked.
Third pair is an example of a very dynamic recording of a jazz trio. Only a few spikes hit peak level. You can play this at high volume and it still sounds clean.
There's nothing like the impact of highly dynamic music. When transients get limited, the peaks get spread over their rise and fall time. Quiet spaces that used to provide contrast for transients get obliterated, and along with them, clarity. Full scale music is vibrant and exciting; compressed music is blocked up and congested.
For an audio sample, I've chosen an HDCD recording of Ravel's "Alborada del Gracioso," performed by the Minnesota Orchestra, Eiji Oue conducting, a Reference Recordings from 1997. You'll hear the sample four times. The first and last are unaltered, straight from the recording, full range. The second is compressed. The third attempts to restore the compressed version by taking it back to the original level. You'll hear the level changes easily, but good head/earphones are best for clearly hearing the finer details lost in compression. Caution : this clip starts off very soft and gets very loud. Set your volume accordingly.
You can see in the histogram the compressed version's blocked up peaks, and how the peaks stay flat after restoring the level. Listening closely between the third and last versions really exposes the losses caused by compression.
<<< full scale >>> <<< compressed >>> <<< restored >>> <<< full scale >>>
Would you rather have the choice of whether or not compression is applied to the music you listen to? Would you like to hear all the expressive nuance artists put into their music? You can have it all, if only.
Speak up. Blog. Tweet. Let recording companies know you're not buying into the Loudness War. Tell 'em you're not happy with music distorted by dynamic compression. Make it known you don't like being cheated and you want your money back. Let your favorite artists know you don't appreciate their music being mutilated. Vote with your wallet. Don't pay for any recording unless the label clearly states, "Dynamic Compression Free."
A well researched document onThe Loudness Wars. And an excellent A/V presentation on the history and future of audio compression, the Loudness War, and why :Peace is Almost Here.
Let's look at what happens with a visual example. The image at right is a full scale image, that is, it has areas of the blackest black, the whitest white, and a wide range of smoothly transitioning shades in between. There is only a relatively small amount of black and a smattering of white spectral highlights. Detail is seen in the shades of gray; absolute black and white define the limits. Below is a histogram of the image showing the distribution of values from 0 
There are no winners in this war. Note how pushing all the tones up to high levels has increased the overall, average brightness, likewise done with the audio signal during compression. Now, you may ask, can the dynamics be restored? Well, we can try. However, a lot of information has been lost in the compression process. After previously differentiated values have been reduced to the same value, there's no way to separate out those lost levels or to recover the lost shading.
