Case Study – Compression

A bit about how I personally use dynamic compression in a live-audio setting.

Please Remember:

The opinions expressed are mine only. These opinions do not necessarily reflect anybody else’s opinions. I do not own, operate, manage, or represent any band, venue, or company that I talk about, unless explicitly noted.

Want to use this image for something else? Great! Click it for the link to a high-res or resolution-independent version.

Generally speaking, I try not to go too far into specifics on this site. I avoid a “do exactly this, then exactly that” approach, because it can cause people to become “button pushers.” Button pushers are people who have memorized a procedure, but don’t actually know how that procedure works. I think we need a lot fewer of those folks in live-audio, and a lot more of people who understand WHY they’re doing something.

Of course, in a “how-to” kind of situation, you do need some specific instructions. Those posts are the exception to the rule.

Anyway.

My avoidance of explicit procedures means that this site has very little in the way of “set this processor this way for this effect” kinds of information. This information does have its place, especially when it’s used as a starting point for creating your own live-sound solutions. Not too long ago, a fellow audio-human needed some help in getting started with compression. What I presented to him was a sort of case-study on what I use compression for, and how I go about setting up the processor to get those results.

Here’s what I said to him, with some clarification and expansion as necessary.

Sledgehammers VS. Paintbrushes

Compression is a kind of processing that has a vast range of uses. Some of those uses are subtle, and some smash you in the face like a frisbee that’s been soaked in something flammable, set alight, and hurled at your nose. Of course, there’s also everything in between those two extremes.

As a creative tool, compression acts all the way from a tiny paintbrush to the heaviest sledgehammer in the shop. For my part, I tend to use compression as a sledgehammer.

The reason that I’m so heavy-handed with compression basically comes down to me working a small room. In a small room, doing something subtle with compression on, say, a snare drum is rarely helpful. The reason is because the snare is usually so loud – even without the PA – that getting that subtle tweak across would require me to swamp the acoustic sound with the PA. That would be REALLY LOUD. Much too loud. The other piece of the small-room puzzle is that the acoustic contribution from the stage tends to wash over even very heavy compression with a large amount of transient content. Because of this, a lot of the compression I end up doing turns into “New York” or “parallel” compression by default. (Parallel compression is a technique where a signal is compressed to some degree, and then mixed together with an uncompressed version of itself. This is usually thought of in a purely electronic way. However, it can also happen with a signal that’s been compressed and reproduced by a PA, but also exists as an uncompressed acoustical event that’s independent of the electronics.) This partial “washing out” of even very heavy compression can let me get away with compression settings that wouldn’t sound very good if there were nothing else to listen to.

Also, there are logistical issues. A good number of small-venue shows are very “on the fly” experiences, where you just don’t have time to explore all the subtle options available for processing a sound. You need to get something workable NOW, and if you have time to refine it later then that’s great.

In another sense, you might say that I view compression less as a sculpting tool and more as a volume-management device. A utility of sorts. Compression, for me, is a hammer that helps me put signals into defined “volume boxes.” For signals like guitar and bass, sticking them into a well defined and not-wildly-changing volume box means that I can find a spot for them in the mix, and then not have to worry too much about level changes. If they get a bit quieter, they are much less likely to get lost, and if they get a bit louder, they probably won’t overwhelm everything. Used across the whole mix, compression lets me basically choose a “don’t exceed” volume level for the PA. I can then push the vocals – hard – into that mix limiter, which helps to keep things intelligible without having the vocal peaks completely flatten the audience.

WARNING REGARDING THE ABOVE: Pushing vocals into an aggressive compressor can be a way to invite feedback problems, because feedback is dependent on gain instead of absolute volume. You can also end up making the stagewash worse, because everything entering a vocal mic (besides vocals) is a kind of “noise.” Running hard into the dynamics processor effectively causes this “noisefloor” to go up. You have to be careful, and if you run into a problem, dropping the vocal level a bit and raising the limiter’s threshold a bit might be necessary. Listen, experiment, work settings against each other, iterate, iterate, iterate…

Setting Up The Sledgehammer

As I said, my tendency is to use compression as a way to stick something into a well-defined space of volume level. That goal is what drives my processor settings.

Attack: Attack time is the how quickly the compressor reduces gain upon signal changes that exceed the threshold. Because my usage for compression is to keep things in a box, I have very little use for slow(er) attack times that allow things to escape the box. Further, I will probably have plenty of “peak” material from the deck anyway, so I don’t have to worry too much. If I notice a problem, I can always act on it. So…I prefer short attack times. As short as possible. Very short attack times can cause distortion, because the compressor acts in a time range that’s less than one wave cycle (which is a pretty good recipe for artifacts.) Even so, I will often set a compressor to the minimum possible attack time – even 0 ms – and live with a touch of crunch here and there. It all depends on what I can get away with.

Release: Release time is how quickly the compressor gain returns toward normal (unity-gain) upon signal changes. I prefer compressors that have an auto-release function which also partially honors a manual release time. With a compressor like that, setting a longer release time with the auto-release engaged means that the auto-release takes proportionally longer to do its thing – while still being program-dependent overall. If I can’t get auto-release, then a manual setting of 50 – 100 ms is about my speed. That’s fast enough to keep things consistent without getting into too many nasty artifacts. (Fifty ms is the time required for one complete cycle of a 20 Hz wave. Letting the compressor release just slowly enough to avoid artifacts at 20 Hz means that you’ll probably be fine everywhere else, because the rest of the audible spectrum cycles faster than the compressor releases.)

Ratio: Ratio is the measure of how much the compressor should ultimately attempt to reduce the output relative to the input. A 1:1 ratio means that the compressor should do nothing except pass signal through its processing path. A 2:1 ratio means that, at or above the chosen threshold, 2 dB greater signal at the input should result in the compressor attempting to reduce the output to a signal that’s only 1 dB “hotter.” Since I’m all about shoving things into boxes, I tend to use pretty extreme ratios…unless it becomes problematic. If a compressor has an infinity:1 setting (where the threshold is the target for maximum output, period) I will almost always try that first.

Seriously, the way I used compression would make lots of other engineers weep. I’m not gonna lie.

Anyway.

Threshold: The threshold is the point at which the compressor’s program-dependent processing is to begin changing from unity-gain to non-unity-gain, assuming that the ratio is not 1:1. In other words, a signal that reaches the threshold and then would otherwise continue to increase should have gain reduction applied. As a signal decreases towards the threshold, gain reduction should be released. If the signal falls below the threshold entirely, then the compressor’s program-dependent processing should return to unity gain – the signal should pass straight through. (I qualify unity-gain with “program-dependent processing” because there are potentially other gain stages in a compressor which can be non-unity and also program invariant. For instance, a compressor’s makeup gain might be non-unity, but it doesn’t vary with the input signal. You manually set it somewhere and leave it there until you decide it needs to be something else.)

Threshold settings are chosen differently in different situations. If I need the compressor to just “ride” a signal a bit, then I’ll try to set the threshold in a place where I see 3 or so dB of gain reduction. If I want the compressor to really squeeze something, then I’ll crank down the threshold until I see 6 dB (or more) on the gain reduction meter. When I’m using a compressor as a “don’t exceed this point” on the main mix, the threshold is set in accordance with a target SPL (Sound Pressure Level) from the PA. However hard I end up hitting the compressor is immaterial unless I start having a problem.

An Important Disclaimer

This case-study was all about showing you how I tend to work. It may or may not work for you, and you should be aware that extreme compression can get an audio-human into BIG trouble in a BIG hurry. This isn’t meant to dissuade you from trying experiments, it’s just said so that you’ll be aware of the risks. I’m used to it all, but I’ll still trip over my own feet once a year or so.

When in doubt, hit “bypass” and let things ride.