Now we’re going to get into more dangerous waters. Bringing up the subject of dynamics compression at a dinner party of audio engineers is like bringing up Mac vs. PC with computer geeks, or XBox vs. Playstation with console gamers.
As I said before, the argument stems from over-use. In my opinion, compression is always needed on a mix, unless it’s incredibly well balanced to begin with, contains a good amount of compression on the individual track recordings, or came from all-digital source instruments. Even then, you’ll probably want to compress somewhat, unless you know your target audience is going to be sitting in a dead silent room, on a couch in front of an expensive stereo system with a nice glass of scotch in hand, quietly focused entirely on listening to the music. And who (besides me) does that?
Ok, so now I’ll take a deep breath, and try to describe what a compressor does. I find that all the technical explanation in the world doesn’t help a lot with using compression, but at least if you know the idea, it can help with figuring out which settings to mess with if you’re not liking what you hear. Compression basically squeezes out the volume differences in a recording. Think of it like a robot hand on the volume knob that is able to quickly turn it down when the sound gets too loud, and back up again when it drops. I’ve always liked this analogy, so I’m going to stick with it. There are two main settings on every compressor, threshold and ratio. The threshold is the volume, in dB, at which the robot hand should start paying attention. Everything below that level should remain unaffected. The ratio tells the robot how hard to twist the volume knob in response to signals above the threshold. As implied, it is a ratio of input to output, so if your ratio is, say, 2:1, it will turn things down enough to make the output only half as much “louder” than the threshold as the input signal. Expressed another way, if a signal goes 2 dB over the threshold, the volume will get turned down to bring it to only 1 dB over the threshold. The higher you set this ratio, the more “squashed” your output will be. In fact, another type of dynamics control called a limiter is really just a compressor with an infinity:1 ratio. The volume gets turned down so that no matter how high the input signal becomes, it’s always turned down enough to not exceed the threshold.
Other features of a compressor…Attack and release times: Briefly, these tell you how quickly the volume knob gets turned in response to volume changes. Peak vs. RMS sensing: This determines whether the robot is reacting to individual spikes in the waveform, or root-mean-square averaging of the signal that is more indicative of “power”. This is useful if your track has a lot of spikes in it, and you find the compressor overreacting in response to a lot of attack transients. You’ll know it when you hear it, trust me. I usually use RMS sensing when remastering, because I know that RMS more closely aligns with human perception of volume, so the result is more natural-sounding, and there aren’t usually any errant transient peaks in something that’s already been mastered that would need fixing. Make-up gain: What we’re talking about here with compressors is technically referred to as “downward compression”. We’re only ever correcting the volume downwards in response to higher input signals. (Upward compression is a whole different thing, done by an expander, and that’s a whole separate discussion) Since we’re turning things down all the time, make-up gain is like a second volume knob that stays fixed at a certain increased amount as overall compensation for the volume reduction the compressor is doing. Look-ahead: Our robot is quickly reacting to input signal as it comes, and making all these volume adjustments for us. But it can only deal with signals once they’ve come in. Real world hardware compressors must work like this, because they can’t see into the future. There are tricks in hardware that can overcome this, which are a little like live network television broadcast censors. They can bleep out all the ‘fucks’ as they come in, but the entire broadcast ends up being delayed a few seconds to allow this to happen. But in digital editing of an already recorded track, we know the future, so we can let software compressor in on it. Then it can react right at a sample level, and not let those transient peaks through because it couldn’t react fast enough. ’nuff said.
Whew. Ok. So let’s do this. Here again is the waveform we’re working with:
At this point, I will state a compressor preference. Sony’s Wave Hammer is like magic for me, and I’ve been using it since somewhere around 2001. In fact, it is pretty much the only reason I keep Sound Forge installed at all times. It has some super-secret-sauce features beyond what a regular compressor has, and a healthy set of presets for dealing with individual instrument tracks, and mastering. I like to fire the thing up, and start with the “Smooth Compression” preset. Here’s what it looks like in action:
This is a capture of it while previewing. I just wanted to point out that the red bar on the right side shows you how much the compressor is actually clamping down on the input signal. All software compressors that you can preview with include a display like this to show you in realtime what they’re actually doing.
So this is a good starting point, but now there’s a bit of tweaking to do. This involves a lot of listening, making slight adjustments, listening again, etc. As I said before, I like RMS sensing better for this kind of thing (Scan Mode: RMS in this case), and I found that the threshold and ratio were a little too aggressive, so I ended up adjusting them. I also checked “Use longer look-ahead”, although I honestly didn’t hear much of a difference.
The “secret sauce” I referred to seems to be a combination of ‘Auto gain compensate’ and ‘Smooth saturation’. You’ll notice that Output gain (which would be the make-up gain I talked about earlier) is set to zero. But still, some amount of volume increase happens to the quieter parts of the track. By the way, this has nothing to do with the Volume Maximizer tab that I’m not showing here. In this preset, it’s all zeroed, and you can bypass it entirely with no effect whatsoever. Anyway, whatever the authors of this tool have done, I’m loving what I hear, so here are the settings I chose before hitting ok:
Here’s what the resulting waveform looked like:
You can see that it doesn’t look like a huge difference, and in fact it really isn’t. Again with the subtle adjustments philosophy. But every section of the track is slightly higher overall, and fatter. It sounds like it too. Here’s the output, if you want to give it a critical listen: https://dl.dropboxusercontent.com/u/4006268/South_Dakota/4.Volume ramp(normalized)(slight hammer).flac
Sorry, no sparkle yet. That’ll be next time…