Audio Processing: Exploring Dynamic Range Compression
Illustrating Compression: Attack and Release
With our final demonstration we hope to shed some light on exactly how different values for the attack and release parameters contribute to gain reduction overall.
Below are two figures, one with all parameters constant except attack and one with all parameters constant except release. Each figure contains seven subplots. The four plots on the left are the source material and then the source material compressed with increasing attack or release time. The three plots on the right show the difference between the source material and compressed version for that parameter set.
Note: All samples were compressed with the FFLog compressor
Showing attack with a drum hit
Figure 1: Drum hit with no compression and then compressed with various attack times
Parameters:
-
Threshold: -15dB
-
Ratio: 5
-
Attack: 0.01, 0.05, 0.1s
-
Release: 0.5 s
Comments:
Notice how with slower attack times (moving down the page) we see less difference between the original signal and the compressed signal. This is because a slower attack time translates to a longer interval over which the compressor applies compression. Because the drum hit is sharp and transient, it isn't above the threshold long enough for the compressor to apply full gain reduction with long attack times.
Showing release with a vocal phrase
Figure 2: Vocal phrase with no compression and then compressed with various release times
Parameters:
-
Threshold: -15dB
-
Ratio: 5
-
Attack: 0.05 s
-
Release: 0.1, 0.5, 1 s
Comments:
This time notice how, as the release time increases we actually get more compression than faster release times. This is because release time is a measure of how long the compressor holds on to the signal and continues to compress after it drops below the threshold. A slower release time will result in a compressor which compresses for longer, thus taking more gain off of the signal. For those that read the technical analysis, this is the release time constant smearing the above threshold output samples across lower level inputs mechanism in action.