I thought it might be nice to take some time to consider the primary tool we use to cope with audio levels. That would be, of course, the audio compressor.
This is a fascinating device. We use it a lot. However, our understanding of it is a little, ah, limited.
The compressor has some fairly tricky controls. Its action can be hard to hear if it's used well. Mostly, it isn't understood all that well by those of us who use it. It has the further characteristic of coming in a range of shapes, sizes and flavors, sometimes described in rather exotic terms that are often hard to understand.
It often is cloaked in retro mysticism (as in "Oh man, that old tube [insert brand name here] compressor does better on everything than any modern compressor. They should never have stopped making it!"). Sometimes, it is an extremely complex device (as in multiband compressors) and sometimes it is pretty tricky as well (as in "look-ahead" compressors).
SO WHAT DOES A COMPRESSOR DO?
A compressor is a device that regulates the gain or level of an audio signal as a function of (usually) the changes in amplitude of that signal, according to a fairly complex set of rules.
Sometimes we use it to prevent overly loud signal peaks from distorting, and sometimes to smooth out the level variations to make a signal (particularly a voice signal) more continuously and easily audible. Sometimes we use it to reduce the overall dynamic range of an audio signal.
All compressors work by sending the incoming audio signal through an active gain stage, usually a voltage-controlled amplifier or its digital equivalent. At the same time, the signal is also sent, in parallel, to a so-called level detector, which studies the signal and converts it into a control voltage, or its digital equivalent.
Said control voltage is manipulated in a variety of clever ways (here's where each compressor gets "its own sound") and then used to regulate the gain of the active gain stage. What could be simpler?
For instance, if the control voltage is inverted — so that as the amplitude of the audio signal gets greater, the control voltage is reduced — the net result will be that as the audio signal gets louder, the gain stage will make it softer, hence "compressing" its dynamic range. Got it? Good.
If only it were that simple …
In actual fact, we want the compressor to do a bunch of other things for us.
First let's consider the Threshold control, which is probably the most important control on any given compressor.
WHAT DOES THE THRESHOLD LEVEL CONTROL DO?
Over the years we've learned that what we really want most is for the compressor to leave the level of the signal alone, except for various embarrassing peak levels that are causing overloads and distortion. Hence, we came up with a control called the "Threshold."
This Threshold control sets the level above which there will be acts of compression occurring. Below that threshold level, there will be no changes in the level of the signal due to the compressor. Got that?
In an analog compressor, the threshold level is usually expressed in dBu (or dBm), and calibrated from something like +20 dBu down to –20 dBu. In the digital realm, threshold level is given in dBFS, from 0 dBFS down to –40 or even –60 dBFS.
Let's study the implications of this for a moment.
What Happens With a High Threshold Level?
Let's set the Threshold control in a digital compressor at –6 dBFS for a moment, to discuss what would happen. When the signal level is at or below –6 dBFS no compression of any sort happens. What comes in goes out unchanged.
However, when the signal level goes above the threshold set at –6 dBFS, compression occurs, and the compressor reduces the level of the signal by some amount (determined by another control called Ratio, which I'll talk about later). This means that as signals approach 0 dBFS they are turned down in level, so hopefully they won't distort.
With this setting, then, very little would happen except when peak signal levels exceeded –6 dBFS, in which case they would be turned down to avoid overloads. Make sense?
What Happens With a Low Threshold Level?
However, if we set the threshold at, say, –20 dBFS, a lot more compression will happen.
First, everything above –20 dBFS (which is probably almost the entire signal trace) will be turned down by the amount determined by the ratio control. So virtually all the audio will be compressed in range, which is quite definitely audible in most cases. In crude or excessive usages this is referred to as "squashing."
So, depending on how the threshold is set, a lot or a little can happen to the signal, from nearly inaudible to extremely audible.
Take a look at Figs. 1–3.
Fig. 1: A hypothetical uncompressed audio signal ranging between approximately –3 dBFS and –16 dBfs
Fig. 1 shows a hypothetical uncompressed audio signal ranging between approximately –3 dBFS and –16 dBFS. The threshold is set at –6 dBFS, but is not active.
Fig. 2 shows the same audio signal, now compressed above the –6 dBFS threshold. Note that there is very little change, except for the peaks; what was –3 dBFS is now –5 dBFS. Very subtle, but it means you could turn up the level by 5 dB without trouble if you wanted.
Fig. 3 shows the same signal, but with a –20 dBFS threshold. Now we've done some serious squashing. What was formerly –3 dBFS is now –12 dBFS and what was formerly –16 dBFS is now –18 dBFS. The dynamic range of the signal was originally 13 dB. It is now 6 dB; dramatically different. If we wished, we could turn it up by 11 dB without overload. It might not sound so good, but we could do it.
Fig. 2: This is the same audio signal in Fig. 1, now compressed above the –6 dBFS threshold.
In a conventional compressor, the only time that gain reduction happens is when the audio signal is above whatever threshold level we've set. When the threshold is set at a high level, little gain reduction happens. When the threshold is set at a low level, a lot of gain reduction happens.
In the latter case, the overall level gets turned down quite a bit. In my previous example, it had been turned down by 9 dB, which is a lot. Take a look at Fig. 4, which shows a composite of Figs. 1 and 3 from above. It shows the heavily limited result of a low threshold applied to a typical fairly loud audio signal. We've limited the gain all right, but we've also made the signal much softer.
That leads us to the use of two other controls on a compressor, Make-Up Gain and Ratio.
Often, our goal in using compression is not to make the signal softer, but to make the signal stay fairly close to some particular level. So, at the output of the compressor, we put in another gain stage with a level control, which allows us to turn the signal that we've squashed back up in level (see Fig. 5).
Fig. 3: This is the same audio signal as in Figs. 1 & 2, but with a –20 dBFS threshold.
Fig. 5 shows several things. First, the squashed signal has been turned up so that now it hovers just below 0 dBFS. This was done by adding 11 dB of make-up gain in the compressor. As a function of doing that, of course, we have also turned up the noise floor of the audio signal by the same amount: 11 dB. This is a polite way of saying that we have reduced the dynamic range of the signal by 11 dB.
So, the benefit of this sort of compression is that we've made the signal level much more stable, and we've made it all pretty loud (and audible). The downside is that we've increased the level of noise, which could be (and often is) annoying. It comes with the territory, as Custer used to say.
The point of make-up gain is to compensate for the amount of gain reduction we've done while squashing the signal, with the negative side effects of increased noise and a certain lack of dynamic expressivity.
SO WHAT IS RATIO?
There is another way to approach this, which is to control the amount of gain reduction that occurs above the threshold level. This is done with the control called Ratio.
The Ratio setting expresses the amount of input amplitude above the threshold that is needed to yield 1 dB output above the threshold. For instance, assume the threshold is at –10 dBFS and the Ratio control is set at 3. This means that an input of –7 dBFS (3 dB over threshold) will result in an output of –9 dBFS (1 dB over threshold). If the input goes up to –4 dBFS (6 dB over threshold), the output will be –8 dBFS (2 dB over threshold). Got it?
Fig. 4: The heavily limited (squashed?) result of a low threshold applied to a typical fairly loud audio signal. We've limited the gain all right, but we've also made the signal much softer.
Suppose the ratio is set at 20, with the threshold still at –10 dBFS. An input of 0 dBFS will result in an output of –9.5 dBFS (0.5 dB over threshold). Usually, any ratio setting greater than 10 is thought of as limiting.
For the illustrations above, I used a ratio of 3.
For solo acoustical musical tracks like vocals, very gentle ratios (maybe 1.5 or 2) are most appropriate. Some¬times though, with a high threshold setting, you'll want a ratio that is also high, particularly if a lot of the program is above threshold, to keep all the overshoots from distorting.
Threshold and Ratio are often used together to manage the dynamics of a given track or mix. You work on the threshold level to get it to a point where it is at or slightly below all of the troublesome levels on the track, while tweaking the ratio to get just the amount of gain reduction you want for effect. In plug-in compressors, you can even automate these two functions, to very gently massage the track as it goes along, variably compressing what needs some help while leaving the rest of the track pristine.
In a compound compressor, there may even be both a limiter function (at a high threshold level) and a gentler compressor function at a lower level. I find this handy to protect myself against overloads while getting some really nice smooth gain management over the meaty musical content of the middle levels of the dynamic range (say, from –8 dBFS down to –25 dBFS).
Fig. 5: This figure shows several things. First, the squashed signal has been turned up so that now it hovers just below 0 dBFS. This was done by adding 11 dB of make-up gain in the compressor. As a function of doing that we have also turned up the noise floor of the audio signal by the same amount: 11 dB. This is a polite way of saying that we have reduced the dynamic range of the signal by 11 dB.
ATTACK AND RELEASE
All compressors work by sending the incoming audio signal through an active gain stage, usually a voltage-controlled amplifier (VCA) or its digital equivalent. At the same time, the signal is also sent (in parallel) to a so-called level detector, which studies the signal and converts its amplitude into a control voltage (or digital equivalent). This, in turn, is used to regulate the gain of the signal via the VCA. That level detector and the resulting control voltage have some pretty tricky aspects, having to do with time.
Three things often happen, and you need to know about them. The first one is a distortion-like sound that can be created by the control voltage being changed too quickly. It's called amplitude modulation. The second one is a change in spectrum, due to the attack and release times emphasizing one portion of the spectrum of the program, or attenuating another portion. The third one is called "pumping," a gasping quality that relates to the level being returned to its uncompressed state too quickly or too often.
All of these have to do with time. In most cases compressor designers have included two special time controls called Attack (to control how rapidly the compressor "attacks" the level-over-threshold to turn it down) and Release (to control how rapidly the compressor stops reducing the gain after the level-over-threshold has gone away).
You'd think, intuitively, that we'd like the compressor to track the "level" of the signal just as quickly as possible, right? That way, we could compress just the offending elements and nothing else. It'd be the most accurate, no?
Unfortunately, the level of the signal is actually the wave trace itself. If we track it very closely, the control voltage will become an audio signal itself (because it goes up and down within the audio range), and cause the voltage-controlled amplifier to also generate an audio signal. This will modulate the actual audio in the low-frequency realm. It sounds just like fairly nasty, low-frequency harmonic distortion.
So, we use two strategies to head off this nasty. Set either the attack or the release control slow enough so the compressor can never modulate in the audio realm. "Slow enough" means a time greater than 50 ms.
Just so you know, 50 ms is the period for a 20 Hz tone, and if we keep things longer than 50 ms, they'll never get into the audio range above 20 Hz. Clever, eh?
If we want a fast attack to catch some sudden spikes of energy, we need to make sure the release is a good bit slower than 50 ms. If we want a fairly fast release, we may want to set the attack slower than 50 ms. As always, use your ears!
Changes in Audio Spectrum
An unintended consequence of compressor use can be an alteration of the balance of the audio spectrum. This happens as a function of the specific program material, and is more of a problem with music than voice.
Often, for example, we'll have extremely loud bass or kick drum signals. They can be (and often are) the loudest component in the program. They can trigger the compressor's gain reduction, turning down the bass, but then (as the release occurs) let the higher frequencies through with little or no attenuation. Occasionally, the behavior will be just the opposite, depending on the settings and the program.
The problems also occur as a function of the attack time and/or the release time changing the envelope (shape of loudness) of individual sounds, and thereby changing their timbre. Solving these problems takes some practice and careful listening to the effect the attack time is having, and then the effect the release time is having.
And remember, it varies with the program material. Heavy metal will suffer differently than will a nice female jazz ballad. But they both will suffer from compression abuse.
Pumping is a real problem in situations where we have both a voice and background noise. We feel the need to limit the level of the voice, to make it consistent and distortion-free. So we compress with a fairly low threshold and large ratio. Also, we use a fast attack, so that the beginning consonants of words don't spit at the listener.
If we use a slow release, once the compressor is invoked, the level stays down, maybe too far. So we speed up the release time, and as a result, during the spaces between words or sentences, the release of the compressor pulls the background noise back up, maybe by 15–20 dB, creating a kind of a sudden sucking "sheeeuppp!" leading into the next word spoken, which then punches the level back down. When this happens over and over, it is fatiguing and annoying.
Compressor designers usually include two special time controls called Attack (to control how rapidly the compressor 'attacks' the level-over-threshold to turn it down), and Release (to control how rapidly the compressor stops reducing the gain after the level-over-threshold has gone away).
The answer here is judicious balancing of attack time and threshold together, against the release time, so that the attack of the compressor is fairly mild and not overdone, while the release is slow enough that the background noise just begins to start pulling up before the next word (and attack) arrive. Some¬times just reducing the total amount of gain reduction (via threshold adjustment) can solve the problem best.
Taken together, the effects of attack and release adjustments suggest a little more about the complexity of compression. Much of what we need to do is more concerned with time than level.
Compressors go back and forth be¬tween the realm where the level change can occur to each individual sound (fast attack and release, which changes the envelope of each sound, a key determinant of timbre) and the realm where we are changing the overall level (which is what happens with a slow release).
Getting so you can hear the difference between compression as envelope and compression as level control is a big step in developing your hearing skills.
To conclude this review of compressor basics, let's consider the nature of the level detector itself, and how it affects the sound quality of the compressor.
THE NATURE OF A LEVEL DETECTOR
The level detector is a circuit that observes the incoming audio signal and derives from it a DC control voltage that will, after manipulation by the Ratio, Make-up Gain, Attack and Release controls, regulate the gain of the voltage-controlled amplifier that is at the heart of the compressor. Exactly how that detector derives such a DC control voltage is an important determinant of how the compressor will regulate gain and how it will sound.
At the heart of such concerns are the concepts of peak, average and RMS levels and their detection. These days, digital compressors often give us a choice between these modalities.
Peak detection derives a control voltage from the highest signal peaks encountered. The detector senses the highest audio level and determines the control voltage from that, usually slowly decaying until another peak is encountered. Such detection concentrates on preventing overloads or clipping, by emphasizing quickly the maximum signal levels encountered.
Average detection is derived from an average of the changing signal level over some time period, usually between 100 ms and 3 seconds. Such detection of amplitude is fairly slow (depending on the signal itself) and gain regulation is comparatively restrained.
RMS (Root Mean Square) detection is derived from the relative power variations of the signal. Power changes as the square of the amplitude, so RMS detection is more volatile over time for a given time constant compared to average detection. It is fairly well correlated to how we humans hear, but it is also fairly complex to compute in real time, so it is not used as much as it might be, particularly in low-cost compressors.
Loudness is a subjective sensation, not a physical value. Our perception of loudness varies as a function of amplitude, frequency, frequency bandwidth and time. It is a complex sensation. It defies direct physical measurement.
With compressors we are attempting to manipulate relative loudness and the behavior of the level detector can have a significant effect on such perception of loudness. In the case of peak detection, a squashing of loudness range occurs for any program with regularly occurring peaks, while for programs with only one peak there is little reduction in loudness except for the few moments following that single peak.
What should be clear from this is that peak detection will have a distinct set of behaviors and effect on loudness, which may not be appropriate, but will be distinctive. At the same time, such peak detection does head off overloads and clipping pretty definitively.
Rupert Neve Designs Portico Series 5043 Compressor/Limiter Duo
In the case of both average and RMS detection systems, we are actually taking an average of the level. Often, we integrate that average over time, so that recent events have a greater effect on gain regulation than do not-so-recent events. But, in any case, such detection is of an average level, not a maximum level. This means overshoots remain possible (governed, of course, by the Attack and Release controls).
The difference between average and RMS is that average values respond to relative amplitude while RMS values respond to the relative power of those amplitudes. That power curve is exponential, so we obtain a different range and nature of behaviors.
Which is the correct one? Neither. They sound different — you gotta use your ears to decide what "sounds best" for you in any given application. Dang!
THE CREST FACTOR
One final complication to all of this is the crest factor, which describes the variable relationship between peak and RMS values.
Compression and limiting both may change the crest factor of a signal, depending on how their time constants are set. At the same time, crest factor is often surprisingly large. For random noise the crest factor is slightly less than 12 dB, meaning that the measured peak level of a pink noise signal will be almost 12 dB greater than the RMS measured level. For some voice signals or certain types of percussive music, a crest factor of 20 dB is reasonable to expect.
Crest factor is not only audible, but it is also a basic determinant of audio character, which is to say that a program with a large crest factor has a distinctly different sound character and sonic and musical meaning than a program with a small crest factor. When we squash the crest factor with a fast-acting limiter, we change something fundamental about that sonic "meaning" of the signal. Not necessarily good.
Compressors and limiters are complex devices. Their action is audible in ways that are often unexpected and sometimes hard to describe. We need them in our work, but we often don't understand all the sonic implications of what they are doing.
One of the key things to keep in mind about a compressor is that we don't normally hear gain reduction as such, unless it is massive. The track may feel warmer, more stable and secure, easier to listen to for reasons that don't seem obvious, and so on. As a general rule, we just don't hear the level itself bobbing up and down.
I personally think this is why compressors are so hard to hear, and to understand.
The behavior of the level detector in each given compressor is a big part of this. And beyond the parameters I have already described, designers can add all sorts of things to manipulate the detector. Unfortunately, such designs and the capability of any given level detector are usually not revealed (except in marketing terms such as "hyperacoustic temporal sensing"), so we are limited to guessing at what they are doing, which can be annoying and difficult.
Such confusions require that we use our ears, which is a good thing, because that's all that our end-users have.
Thanks for listening. And thanks to Eddy Bogh Brixen ("Audio Metering," Broadcast Publishing) for some really useful information for this article.
This series on the basics of compressors was published in TV Technology magazine. Reach Dave Moulton via his Web site moultonlabs.com.