If you’re just tuning in, we’re in the middle of a three-part series about compression in the recording industry and how that applies to classical music. Part one discussed why this is necessary and why we should be advocating for this to happen on our recordings. Today, part two discusses the technical side of this problem by explaining what compression is and how it is used. Next week we’ll discuss the kind of pushback I’ve seen from those in the recording industry on this topic.
What is compression?
A compressor (the tool that engineers use to add compression to a signal) falls into a family of effects processors called dynamic effects. All this really means is that a compressor makes changes to the relative volume of a signal. Other dynamic effects are gating, expanding, and limiting. They’re all basically related and work around the same principals. Simply put, compression is a process that engineers use to limit the dynamic range of a signal. It makes things that are loud a little (or a lot) quieter, and makes things that are quiet a little (or a lot) louder. It does this by automatically sensing the level of a signal, reducing that level (attenuation) if it goes beyond a certain point (threshold) and then turning the overall volume of a signal back up to compensate for this reduction (output gain).
A compressor generally has three very important controls: threshold, ratio, and gain makeup (also sometimes called “out gain” and various other iterations). There will almost always also be several other controls (attack, release, knee etc.), but these three are where the main business of compression gets done.
Compression happens in a three step process that is embodied by these controls. The threshold control sets the point at which gain reduction begins to occur. Let’s break that down a little bit. The image below is a wavefrom representation of an audio signal. The horizontal axis represents time, and the vertical represents volume. So, what we’re looking at is how the loudness of a signal changes over time.
You can see that this signal starts very quietly, and then something explosively loud happens before a more moderate volume level takes over. The first step in compression involves setting the threshold. In the image below, I have drawn a line at about 0.25 dB. Let’s say that this is where we set our threshold.
This means that any of the signal that goes beyond the red line (is louder than 0.25 dB) will cause the compressor to apply some gain reduction (turn down the volume).
This is where the ratio control comes in. The ratio determines how much gain reduction will be applied once a signal goes beyond the threshold level. This is expressed as a ratio of pre-attenuation to post-attenuation decibels. The control in the picture above has markings like 2:1, 4:1 etc. This means that if the ratio knob is set to 2:1, any signal that is 2 dB above the threshold will be attenuated until it is 1 dB above the threshold; any signal that is 4 dB above the threshold will be attenuated until it is 2 dB above the threshold and any signal that is 1 dB above the threshold will be attenuated to0.5 dB above the threshold. The same is true of 4:1: 4 dB in equals 1 dB out; 1 dB in equals 0.25 dB out and so on.
For demonstrations sake, let’s apply a pretty significant ratio, about 5:1, to our waveform. Look what happens.
The very loud thing is now significantly quieter, but the volume of the other elements remains the same!
The last step is gain compensation. The idea behind this is that by turning down the loud parts of a signal, we lose a certain amount of overall volume, so we turn the signal back up post attenuation to compensate for that loss of volume. In this case, I’ll boost the signal back up so that the very loud part is about the same volume it was pre-attenuation. You might be asking why I would turn it back up after I just turned it down. Bear with me for just a moment…
After everything, the overall effect was to compress the dynamic range so that the quiet parts are actually louder, without losing any volume in the loud parts!
You might now be asking why we didn’t just turn the overall volume up. The simple answer is that we weren’t able to do that because of that very loud thing wouldn’t permit it. If we had turned up the overall volume in an attempt to make the quiet parts louder, that loud thing would have gotten louder to the point that it would overload and distort (or hurt someone’s ears, or blow someone’s speakers etc.). This is why this doesn’t work well in the car, turning the overall volume up to make the quiet parts audible makes the loud parts unmanageable.
This is basically how compression works. There’s a lot of nuance that come out of this process, but that isn’t necessary to discuss here. It’s also worth mentioning that the images I use above used relatively extreme settings so that you would be able to actually see the different processes at work. More common compressor settings, those for which I actually advocate, would be more subtle and more difficult to see in action.
As mentioned above, compressors also come in a lot of different flavors. One of the most important ones in this discussion is called a limiter. This is usually a tool that’s used in the mastering stage of the recording process. This is the final stage before audio can be released commercially. Mastering is done to put a final polish on the recording using eq, and to make sure that it is loud enough, and that all the songs on an album are the same level. This is where the limiter comes in. A limiter is like a compressor with its ratio permanently set at :1. What this means is that, no matter how far above the threshold the music goes, the limiter will turn it down until it is AT that threshold. Imagine the scene in The Lord of the Rings when Gandalf shouts at the Balrok “YOU SHALL NOT PASS!” That’s what a limiter does to an audio signal. This means that you can turn the signal going in to the limiter UP and what comes out will still come out at the threshold level, so the result is a signal that is perceptually significantly louder than it was before limiting. This is because, theoretically, a limiter provides you with the ability to make the quietest moment in your music exactly the same level as the loudest moment.
This isn’t actually what I’m advocating for. There needs to be some amount of dynamic contrast in classical music. Crushing a recording to death so that there isn’t any actual contrast at all makes it unpleasant and tiring to listen to. But if the dynamic range of a recording is too large, it’s essentially impossible to listen to. What needs to happen is that the quiet end needs to be brought up to a reasonable level for the most common listening situations, without the loud end becoming oppressive. Compression is the only way to do this.