go up a level

Good and bad compression

The creation of music can be depicted by the following diagram: 

                               ---------
instrument_1 --> effects_1 --> |       |
instrument_2 --> effects_2 --> | mixer | --> music
...                            |       |
instrument_N --> effects_N --> |       |
                               ---------
                    fig.1

The diagram shows how instruments generate their sounds. These sounds are processed by effects, summed up in a mixer, and, ultimately, music is produced.

Dynamic compression can be used as an effect, among others. The following figure shows oscillograms of the same soundtrack before and after compression.

sound oscillograms before and after compression

To make the difference between the soundtracks more visible, we normalize both soundtracks by amplitude:

sound oscillograms before and after compression

Now you can easily notice the following:

p.1  Compression reduces the difference between the loudnesses of the quietest and loudest sounds, i.e. the loudness is equalized throughout the soundtrack.

Considering that loudness is associated in the brain with the distance to the sound source: further - quieter, closer - louder, then we can say that the stronger the compression, the closer the sound sources are to each other. Therefore, it is appropriate to introduce the concept of the volume of the sound scene. Compression reduces this volume.

p.2  Compression increases the average loudness of a soundtrack (or rather, the average loudness relative to its maximum): a normalized compressed soundtrack sounds louder than a normalized original one. A visual comparison of the soundtracks suggests that they can be compared not only in terms of quieter/louder, but also thinner/denser, weaker/more powerful.

Compression itself is neither good nor bad. It becomes good or bad depending on the application.

Example 1. Let's assume that in the diagram in fig.1, instrument_i creates a sound that is too uneven in loudness (this could be, for example, an undesirable fluctuation in the loudness of the voice depending on the distance between the mouth and the microphone). In this case, the loudness can be leveled using compression.

Example 2. Let's assume that instrument_i sounds too weak or too thin against the background of other instruments and this cannot be eliminated by simply increasing the loudness, then compression can also be applied.

The volume of the sound scene in both of these cases will not suffer, because the compression is applied to one instrument, which already occupies one point on the scene, or if there is volume, then it is very small compared to the scene.

Both of these examples are examples of good compression. Now let's move on to bad compression.

Let's compare the oscillograms of the composition "Daft Punk - Alive" released on the release "va - Soma Quality Recordings - 1994 (CD)" and on the release "Daft Punk - Homework - 1997 (CD)".

Daft_Punk - Alive

The first oscillogram does not arouse any suspicions, but the second one does - because it is too flat, like a brick. Let's recall fig.1 and ask ourselves: is it possible in nature, in natural conditions, that such a flat oscillogram would be obtained when summing several sound sources? No. Hence, after summing, i.e. after the mixer, the music was subjected to some additional processing, and such processing that contradicts nature. What kind of processing is this? - it's not hard to guess - it's compression: 

                               ---------
instrument_1 --> effects_1 --> |       |
instrument_2 --> effects_2 --> | mixer | --> music --> compression --> compressed
...                            |       |                               music
instrument_N --> effects_N --> |       |
                               ---------

In this case, the instruments are not compressed individually, but all at once. Compression forces them to be located closer to each other (see p.1). If before compression the instruments were located all over the scene, now they are lined up in front of the listener, the volume of the sound scene has decreased, the sound scene has turned into a sound wall. In addition, the sound has now become unnaturally loud (see p.2).

This is an example of bad compression.

Since the late 90s, almost all published music has been crippled by compression - it looks like a brick and sounds accordingly - unnaturally flat and monstrously loud. In the late 90s, thanks to technical progress, the production and consumption of music became cheaper and easier. Too many strangers began to compose and listen to music. They cannot appreciate such a subtle quality of sound as the volume of the sound scene, but they can appreciate loudness. So why not sacrifice the volume, which no one understands, for the sake of the loudness, which everyone understands?