Compression
In audio signal processing, compression (dynamic range compression) is the operation of reducing the levels of loud sounds in an audio signal. This technique is widespread in music production, mixing, and mastering.
Compression at a surface level is fairly simple to understand, but can be difficult to hear when applied subtly. When the input to a compressor passes a threshold (in dB), the compressor activates. How quickly the compressor activates is called the attack time. How quickly the compressor deactivates after the input signal drops below the threshold is called the release time.
When the compressor activates, it takes the input signal and reduces its level by a set amount. This amount is defined by the compressor’s ratio. If an input signal exceeds the threshold by 2 dB, and the ratio is set as 2:1, then the output signal from the compressor will be 1 dB above the threshold.
Now, you may already have noticed: the signal still ends up being above the threshold! If we were to increase the ratio to 10:1, then a signal crossing the threshold at 2 dB above would be reduced to 0.20 dB above, which is hardly audible. If we increased the ratio to 50:1, the output level would be 0.04 dB above the threshold. So on and so forth. If the ratio were “infinity”, then the output level would be equal to the threshold.
Colloquially, a compressor with a ratio of 20:1 or higher is equivalent to a limiter. In practice, compressors do not have a ratio this high as the result sounds more “artificial”.
Alright, most of what I wrote above is only true if the compressor is using a hard knee. With a hard knee, compression “turns on” as soon as the level passes the threshold. (The attack time must be considered first.) The compressor uses the “full” amount of compression, as defined by the ratio. That is, regardless of the input level, the same ratio will be used whenever the level passes the threshold.
If the compressor is using a soft knee, then the amount of compression will depend on how close the input level is to the threshold. In many cases, compression will actually start being applied before the input level passes the threshold, albeit at a lower ratio. For example we could have a compressor that behaves as such:
- Assume a ratio of 10:1 and a threshold of -10 dB
- When the input level is at -15 dB, a ratio of 2:1 is used
- When the input level is at -12.5 dB, a ratio of 4:1 is used
- When the input level is at -10 dB, a ratio of 8:1 is used
- When the input level is at -8 dB, a ratio of 10:1 is used
So the compressor doesn’t apply its full potential until a point actually beyond the threshold, not just the threshold itself. (These values were made up: refer to the compressor’s manual!)