Reference: What Is The Compressor Thingy Doing - Anyway

Unless you are a recording musician, you probably are unaware of how much compression used in everyday audio. This is the single most important tool in any engineer’s bag of tricks, and it separates the pros from the wannabes.

Unlike other effects (EQ, distortion, reverb, delay, flange, chorus, etc), the compressor does not add coloration or extra material to the audio signal. It simply changes the dynamic properties of the sound. The ironic thing is that if it is used correctly, most people do not know it’s even there. However, when used incorrectly (or not used at all) your work screams “amateur.” Believe it or not, EVERYONE notices when it is absent or mis-used, even though they would not be able tell you what was missing.

A compressor flattens the dynamic range of a sound. Think of it as an automatic volume control, and you are not far off: it is as if you have an “invisible assistant” with his hand on the volume knob, and his only job is to keep the sound coming loud enough for you to hear well all times. How many times have you watched a movie where there is yelling, screaming, and gunfire, then they cut to a scene where some injured schlup whispers his last words, and you could hear it as well as if he were whispering in your ear?

Compression!

Not convinced? Try this experiment: Have a friend standing about 6 feet away from you scream as loud as possible, fire a gun, and then whisper. I can guarantee you won’t hear the whisper.

Compres

ors use a ratio system to determine how much they flatten the dynamic range. A compression ratio of 1:1 means that for every one decibel of input, you get one decibel of output. Here is a chart showing with some reference volumes:

Figure 1

It is interesting to note that the decibel scale is logarithmic: each 3dB increase is equivalent to doubling the volume. If you compare 60dB (Speech at a foot away) and 120dB (Front row at a rock concert), the first response is typically “There’s no WAY a rock concert is only twice as loud as someone talking!” It isn’t. It’s 20 times as loud.

Looking at the chart above, the Input sound is the vertical axis on the Left side, and the Output (what you hear) is the bottom. At a 1:1 ratio, a 40dB input sounds like 40dB, and 120dB sounds like 120db. Make sense so far?

Now let’s look at what happens when we compress it.

Figure 2

This is our same chart, but this time we are putting a 2:1 compressor between the Input source and ourselves. Remember, a 2:1 ratio means that for every 2 decibels in, we get 1 decibel out.

At a 2:1 ratio, a 120dB noise (our front row rock concert) becomes a 60dB noise (conversation at a foot away). Again, there is the logarithmic nature of audio coming to fool us again. We think 2:1 means that it is reducing the volume by a half, but there is really much more going on than that.

Looking at the chart we can see that the dynamic range is reduced through the entire range of sound: everything from a barely audible 0dB to the choking sensation of your chest collapsing at 150db has been reduced to a manageable 0db to 75db.

At a 2:1 ratio: a 40dB input is 20dB, and 120db is now a quiet 60dB.

This is only half of the story, because what we really want to do is make the quiet sounds louder. Right now, we’ve succeeded only in making the loud sounds much more quiet. This isn’t really that useful, so the output of a compressor always has a Gain control that brings the output back up to a “normal” level.

Here is our 2:1 compressor with the gain turned up so the loudest sound (our chest crushing 150db noise) is still 150db on the output:

Figure 3

If you compare the compressed to uncompressed sound, this is the result:

- The loudest uncompressed sound is exactly that same volume after compression,

- The softest uncompressed sound is quite loud after compression

By bringing the Gain up after compression, we have indeed succeeded in making the soft sounds louder. A quiet conversation at 40db is now somewhere between noisy traffic and a New York subway terminal. Nice!

Of course, there is no such thing as a free ride, perpetual motion, free beer, or perfect audio. While we did get our signal into a more manageable range, we also introduced a lot of noise into our formerly pristine recording. Every tiny background noise in the input source has been boosted so that things that were just at the threshold of hearing are now at a volume level louder than conversational speech.

This means that previously “invisible” sounds (say someone coughing in the distance, people breathing, walking, setting a can of soda on a table) are all crystal clear. (This is why they call out “Quiet on the set!” before they roll film!) For music production, the same problem manifests itself in different ways.

Commercial songs are recorded in expensive studios for a reason. Yes, they have top notch gear, but they are also incredibly quiet. Not only are environmental noises reduced past the threshold of hearing, but internal systems like air conditioning and plumbing are also designed to be quiet as possible. This is absolutely necessary because modern popular music is highly compressed. Most individual instrument tracks are processed, and the resulting mix often ends up being run through at least one stage of compression during mastering.

If you isolate a single drum microphone and listen to the track after raw compression is applied, you will hear amazing things. Squeaky foot pedals, noise from the cymbal stands, grunting (you think I’m kidding??), and worst of all: other drums that suddenly sound almost as loud as the one you wanted to isolate.

Take that same track recorded in the typical home environment, and it gets even more interesting. Between beats of this incredibly loud floor tom, you will hear the air conditioner kick in, the TV in the other room, whirring of computer fans, and worst of all: traffic outside.

Suddenly you are faced with the prospect of having given yourself a lot more work trying to ~remove~ these elements from your track, and that’s just one track. The typical song will take 16 or more tracks, and the noise on all of them will add up. Is it worth it? Just for kicks you turn off the compressor and listen again: the background noises are once again insignificant, but all those undertones that make the instrument ~jump~ out bigger than life are gone too.

The solution is setting a Threshold level before the compression kicks in. But before we get into that, we need to take a closer look at how the decibel level meters on a recording console differ from the absolute decibel level used in the physical world.

DECIBELS AND DECIBELS

So far we have been referencing dB as a physical measurement of how loud a sound is. In this realm 0dB is the point where you can just start to hear a sound, and goes up from there. In the recording studio, this scale is inverted.

In audio gear, all meters and equipment are set so that 0dB references the MAXIMUM loudness of a signal, and goes down from there. To get the best signal-to-noise ratio when recording, it is necessary to record a track as loud as possible without overloading the recording medium. Back in the old analog days, this was a no-brainer: if you recorded a track softly and then tried to play it back at a high volume, you’d have a ton of tape hiss.

To make it easy to determine how loud to record a track, the equipment is calibrated so that a constant signal of 0dB represents the loudest signal that the equipment can handle. When recording, the goal is to get the signal level as close to 0dB without going over.

Current CD’s play back 16 bit audio. Without going into a lot of math to explain it, this translates to a maximum signal-to-noise ratio of 96dB. (When CD’s were introduced to the public in the early 80’s, this was a huge improvement over the 60dB that a phonographs were capable of!) Therefore, most audio gear currently in studios have meters that run from -92dB to +3dB. (The +3db is there for those brief peaks over 0dB, so the equipment doesn’t distort.)

THRESHOLD AND KNEE

OK, so let’s throw up a graph of our good ole 2:1 compression with Gain from Figure 3, but I am going to re-label the scale so that it matches what you will see in the studio:

Figure 4

In this example, we have the Threshold set at the minimum, which means the entire signal is processed. At a 2:1 ratio, for every decibel 2 decibels OVER the Threshold, we increase the output 1 decibel.

Let’s assume we are trying to set the compression on a Kick Drum on a standard drum kit. Even at a modest 2:1 ratio, you will notice immediately that undesirable background noises have also been increased greatly. Particularly, bleedover from other drums is very evident after compression.

A lot of this will be solved by setting the Threshold high enough that the compressor doesn’t start to work until the signal rises above that level. We’ve either watched the meters (or opened the track in an audio editor) and have determined that none of the background noise in this track ever rises above -30dB. Setting the Threshold at -30dB yields a chart as show here:

Figure 5

This is a little better. Background noises have dropped back to decent levels, and if we have a good quality recording in either a studio or very quiet home studio, we might be able to stop there. Note that the background noises are still louder than what they would be without compression. (For instance, the -40dB Bleedover is about -25dB or so). This is due to the Gain that we have added to the output of the compressor to bring the 0dB Input signal back up to 0dB on the output.

If the background noise is still a problem, you have a few options:

- Reduce the compression ratio,

- Re-record the track using microphone and isolation techniques to minimize the noises,

- Use a Noise Gate or Expander to reduce signals below the Threshold level to much lower levels.

Looking at Figure 5, you’ll notice that the threshold level where the two lines intersect is a “hard” intersection. This means as soon as audio crosses that line it goes from one amplification factor to a totally different one. On subtle compression ratios, say 3:1 and lower, this isn’t very noticeable. On higher ratios, and/or certain types of program material, it creates a very harsh sound. This is where the Knee control comes in.

The Knee basically sets a gentle curve between the two amplification lines. Depending on the compressor, your options will vary. Some only have an option of a “Hard Knee / Soft Knee” which means either: A) the Knee is off and you have a hard intersection like in Fig. 5 (Hard Knee) or, B) the Knee is on and there is a curve to ease the transition from one amplification line to the other.

Some units will allow you to specify the amount of knee yourself: the greater the Knee, the wider the curve. The Knee is a very subtle control, and adjustment of this parameter is totally a matter of taste, and what sounds best with the compression and material you are working with.

ATTACK AND RELEASE

The compression ratio and threshold allow you to control the overall volume of the signal. Other than the ability to bring up sounds that are in the background, they really do not change the character of the sound you are working on.

The Attack and Release controls, however, will have a huge impact on the sound. Both the Attack and Release controls are time-based, which means their settings are measured in milliseconds.

This is the short definition of what the controls do:

Attack: The number of milliseconds before the compressor reacts to a signal once it exceeds the Threshold.

Release: The number of milliseconds it takes to return the compressor unity gain (1:1) after the signal falls below the Threshold.

Using our “invisible assistant” analogy from earlier, the Attack time is how long it takes him to react to a change in volume before he starts turning the volume knob: The longer the Attack time, the slower his reaction.

Let’s say we are watching our action movie again, and our trusty assistance is sitting there with his hand on the volume knob. We are watching a quiet, tense scene where there is a bomb counting down the last few seconds, and a sweaty guy is sitting there with a pair of wire cutters. Our assistant has the volume cranked to 11 on a scale of 1 to 10, because we’ve ~got~ to hear the little tick of the clock, hear the guys nervous breathing, and hear the drop of sweat as it falls from the guy’s forehead onto the table.

Oops, surprise! The drop of sweat shorted out something and the bomb exploded 3 seconds before we thought it would. If our trusty assistant has a slow reaction time, our Dolby Digital Certified High-Definition Theatre Surround Sound system with 50,000 watts of body crushing power will likely blow out every speaker, window, and eardrum within a 200ft radius because the volume was still cranked to 11 when the explosion happened.

In this case, we want our assistant to be supernaturally quick, and snap that volume back down to a reasonable level a split second after the beginning of the explosion occurs. That way we get to hear the ticking, hear the breathing, hear the sweat droplet, and experience that fraction of a second of oh-my-God-what-happened intense volume surge that makes us jump when the explosion first starts. Then, we’ll get to hear the rest of the explosion and falling debris at a comfortable volume that won’t break or damage anything, and won’t cause the neighbors to call the police.

This scenario plays out just the same when compressing an audio track. Here is the plot of a single snare drum strike:

Figure 6

Each of the vertical grid lines represents 10mS of time, and this entire sample is only about 210mS long: just shy of a quarter of a second. You can see that the first 15mS or so is the most intense, when the stick strikes the drum head. The rest of it is the decay from that initial hit.

We really want to bring out the tone of the snare, so we put a 4:1 ratio on it so that spring rattle during the decay comes up nicely. You set up the Threshold to about -30dB so the compressor isn’t working on all the other noises coming through. Now it’s time to adjust the attack.

If you haven’t done it before, it’s really instructional to sweep the Attack while the track is playing to listen to how it changes the character. At the minimum Attack setting (mine are digital, so they can go down to 0mS), you lose all of the punch of the initial stick hit. As soon as the signal starts to get loud, the compressor kicks in and squashes it back down. In this case the entire snare sound changes to a very even snare buzz / rattle, which is great for techno or dance music.

But, I like rock, so I need to hear some stick when the drummer hits a drum.

Moving out to about 15mS or so, it sounds nice. We have a solid of the initial hit, because it lets that first 15mS through untouched. Then the compressor starts doing its work, and brings up the rattle and buzz of the snare. At 50mS and higher, the compressor isn’t doing much for us anymore because we’ve let most of the signal through untouched.

Setting the Release time is a different affair. Many engineers rely on the tempo to determine the release time. The thinking behind this is that you want the compressor to just come back to unity (1:1) when it is triggered again, so that it “breathes” with the music.

To do this, play the track and just watch your meters as you slowly increase the Release time. You will hit a point where the compressor is still active when it is triggered again, back it off to the point just before that happens. Done correctly, the compressor will breathe with the music, and add its own character to the sound.

From here, it’s time to sit down at the board and start playing! Once you master the compressor, your mixing will start sounding much better!

Lonnie R. West is a Software Engineer and freelance writer and musician.
http://www.zippybackflash.com

Full article complete with graphics can be found at:
zippybackflash.com/blog/?p=20