Dynamic range compression (DRC), is one of the most widely used audio effects, yet its still misunderstood with very little formal knowledge or analysis of compressor techniques. In the article Intelligent Multitrack Dynamic Range Compression, Zheng Ma, Brecht De Man, Pedro Pestana, Dawn Black and Joshua D. Reiss talk about several different approaches to digital dynamic range compressor design. They have created a list of 6 assumptions that are common and considered to be the best compression tactics that I will list then break down respectively.
1. A signal with a high degree of level fluctuations should have more compression.
2. A signal with more low frequency content should have more compression.
3. Attack and release time should be dependent on the transient nature of the signal.
4. Knee width should depend on the amount of compression applied.
5. Make-up gain should be set so that output loudness equals input loudness.
6. There is a maximum and optimal amount of DRC that depends on sound source features.
In a survey of a handful of professional mix engineers, most implied the main reason to apply DRC is to “stabilize erratic loudness range”. Meaning compressing tracks with high note-to-note level variations so their levels remain more consistent, IE vocals or drums. Dynamic features used such as EBU loudness and dynamic spread are used to measure level fluctuation, but have been found to be inaccurate according to subjective loudness from person to person. The authors suggest to combat this by finding the crest factor by calculating the peak amplitude divided by its RMS value to be a more accurate measurement of the dynamic range.
Based on an analysis of multiple mixes, it has been proven that compression typically takes place whenever headroom is at stake and the low-end is critical. To find the degree of frequency dependence and low-end sensitivity of DRC it’s suggested to utilize spectral features of the source audio signal. Exploring its spectral spread, brightness and spectral centroid.
Assumption 3 is something that may seem easier said than done, utilize your ears and play with the attack and release times to work with the sound of the instruments and the tempo of the song. Some instruments are easier than others, percussive instruments being the most obvious, needing short attack and decay times, while other tracks depend on the play style and instrumentation. Some compressors even offer auto features that read and predict measurements of the audio’s peaks and RMS levels. If you do not have this feature you could be playing with attack times ranging from 5ms to 250ms, and decays times ranging from 5ms to 3,000ms, which can sound daunting but again your ears are your best tool in the audio world. Always trust what you hear.
A soft knee will enable a smoother transition between compressed and non-compressed signals and will yield a more transparent sound. The knee’s width should be configured in according to the estimated amount of compression being used by comparing the relationship between threshold and ratio. Again, this is not using your eyes and numbers, they can help find the general area but by using your ears and listening for the right amount is key.
Assumption 5 is something that I used to commonly make a lot of mistakes in. Matching your make up gain from your plugin with your bypassed audio’s levels is important. Do not mix within plugin parameters. However, lately a lot of DRC products commonly use an automatic make-up gain based on average control-voltage. This can sometimes produce a perceived loudness variation so isn’t suggested. Subjective evaluation comparatively, makes a better tool in making better decisions in setting the make-up gain.
Lastly, and arguably most important, is finding the spot where you get the right amount of compression without destroying the original audio. It’s about forming a discreet separation between peaks and steady states within the waveforms, and allowing the steady states to have variability of the knee width to accommodate for the peaks. Its a fine line to walk along, but again, like all rules in audio it comes down to what sounds right. Trust in what you’re hearing and make the choices based on your instincts.