Audio Compression (June 2018)
Audio compression (not to be confused with audio file compression, a process for reducing file size) alters a track’s waveform by decreasing a signal’s amplitude by varying amounts depending on its original amplitude: high-amplitude waves get a lot of attenuation while low-amplitude waves get little, if any. This is frequently done during recording sessions to avoid “pegging the meters”—capturing sound so loud that it hits the maximum that the recording system can tolerate and results in distortion. Audio compression can also function by increasing low-amplitude waveforms while leaving higher-amplitude waveforms relatively intact, and I will focus on this application for the purposes of this article. Intuitively, the net effect is to decrease the dynamic range of a recording (the difference between the recording’s loudest portion and its quietest). A more technically correct description is that it decreases the recording’s “crest factor,” which is the difference between the track’s peak amplitude and the root-mean-square (RMS) “average” of the entire track.
Beyond making everything louder, compression can be a very useful tool to “normalize” volume among several tracks that will comprise a compilation or playlist so listeners don’t have to continually fumble with the volume control. It’s also useful in adjusting the character of separate instruments or vocals in producing a final mix, or for addressing specific notes or passages that are too loud or too soft. Here’s an example of the latter.
Figure 1 shows a stereo waveform of a classical piece featuring a violin with very soft accompaniment, recorded from a 78, and intended for inclusion in a compilation. Note that the amplitude is subdued, i.e., the peaks of the blue waveforms hover around the center of the displays rather than extending further upward and downward. Left untouched, this track would be noticeably quieter than the other tracks on the compilation. Another thing to note is that there’s a spike about 40% of the way through the track. While the simple solution to the quietness problem would have been to increase the amplitude of the entire track, doing so would push that spike beyond the maximum amplitude that a digital system can represent, resulting in audible distortion.
Figure 1. Uncompressed audio waveform.
The music in this track has uniform volume, and the spike represents an unusually heavy violin bow stroke that, while not objectionable to the listener, is a bit out of character with the rest of the music. Consequently, applying compression increases the amplitude of the track except for the spike. In Figure 2, you will note the peaks of the blue waveforms are larger than they appeared in Figure 1, reflecting greater amplitude. The spike is still present, but since it received very little amplitude increase from the compression, it still falls safely within the digital system’s limits.
Figure 2. Compressed audio waveform.
I describe what has occurred in commercial music as audio compression became “weaponized” in The Loudness War.
Back to Paul’s Blog & Contents