For the last three installments this column has been dedicated to the frequency-specific elements of recording and mixing through the use of eq. Now it’s time to consider the amplitude domain and how it relates to time, timbre, and space in the mix. A special set of audio tools is set-aside for just such considerations… dynamics processors. These include compressors, limiters, expanders, gates, and frequency dependant effects such as de-essers and pop reducers.
The compressor is one of the most common types of dynamics processors. A compressor is a device that automatically reduces the gain of an audio signal when it exceeds a specified level, called a threshold. By lowering the gain whenever the signal “gets too loud”, the dynamic range (the range from softest to loudest) of the audio is decreased.
One way, for us musician-types, to understand this effect is to consider it in musical terms; music that ranges from pp to ff (musical expressions of loudness) can be compressed to a range of pp to mf. But it’s not enough to just knock down the louder instances of a signal and end up with a piece of audio that never gets louder than a given, moderate level. For audio recording purposes it is best to bring the final level back up to a point that will maximize the signal to noise ratio and be more appropriate for audio playback systems. When this is done, the entire range is then shifted upward to become mp to ff (or even mf to fff).
There are several main reasons to reduce dynamic range by use of compression:
- to make the audio louder overall
- to increase the signal-to-noise ratio of a recording
- to automatically reduce differences in performance levels which are distracting or make a track hard to balance in the mix
But volume adjustments are not the only function of compressors – they can also be used to change the timbre (tone color) of audio. Since compressors vary level over time, they can be set to modify the envelope (dynamic contour) of a sound… sometimes drastically. This is one of the fundamental principles underlying people’s perception of timbre. In other words, a compressor can be used to aid in balancing a guitar part in the mix, or to change the very sound of the guitar part. We’ll explore exactly how this is accomplished in TCRM 18.
How Compressors Work
Compressors are built according to a number of different technologies (and software algorithms) that can affect the particular flavor of a compressor’s sound. Despite this, the basic effect and functions are common to all.
Threshold and input gain
The threshold is the level above which the compressor will begin its gain reduction. On many compressors, it is assignable using one of the various dB scales, depending on whether it is analog or digital, and at the discretion of the manufacturer.
The reliance on threshold numbers, while useful under certain conditions, often leads to one of the most common misconceptions about compressor settings: that the threshold setting has some absolute meaning. It does not. The threshold is a relative function. Over the years I have heard many recordists say things to the effect of, “I get the best sound on vocals by X, Y, Z,… and with the threshold set to -12.” What they don’t seem to realize, is that this magic number is practically useless without a much more precise understanding of the nature of the input signal.
The input level is the most fundamental of these concerns. A threshold set at -12dBFS will do nothing if the input level is too low. By turning up the input, that threshold can be exceeded and compression can begin. On the other hand, if the level is too high, the compression may be too great.
For this reason, the threshold is not an assignable parameter on a number of compressor designs. It is fixed. Since threshold is relative to input gain, they rely on gain adjustments to trigger compression. A separate control for gain is found on some models (whether threshold is assignable or not), while others rely solely on other means to adjust the level before it is sent to the input of the compressor. Channel trim pots, fader levels, send levels, and other bus output levels can be used in various situations to accomplish this task.
Compression ratio is the next aspect of compression to consider. This is a description of the amount of gain reduction. It is expressed by the amount (in dB) of input above the threshold that will cause the output to be 1 dB above the threshold. For example, a 3:1 compression ratio would mean that a signal exceeding the threshold level by 3 dB would be output at only 1 dB above the threshold. This also means that a signal exceeding the threshold by 9 dB would result in an output of 3 dB above the threshold. If that were based on a threshold of -10dBfs in a DAW, a signal input at -1 dBfs input would come out at -7 dBfs, a reduction of 6dB.
Attack time and knee
Due to the extremely diverse nature of the way various sound sources initially produce energy they exhibit differing attack characteristics.
Sounds can start out in many ways: hard, soft, fast, slowly swelling up, etc. - these are called attack characteristics.The earliest parts of a sound, called transients, are an important part of defining the character of a sound. Since they are sometimes also of high energy, a fast-acting compressor can partially truncate them (cut them off). For this reason, you can adjust how quickly a compressor begins to reduce gain once the threshold has been exceeded. This setting is called the attack time and is usually expressed in milliseconds.
Some compressors also include an “auto” mode, which automatically adjusts attack time based on aspects of the incoming signal. It is a mistake to rely on this too heavily. What the machine wants to automatically do to your sounds, and what you want done may not always be the same… unless you’re a robot who “thinks” just like your compressor. Also keep in mind that, when auto mode is selected, the manually selected attack time is meaningless.
To create what some consider a more “natural” sounding compression scheme, some manufacturers have included variable-slope attack response curves called knees or contours. Here, instead of reducing gain by the desired ratio immediately upon breaking the threshold, full reduction is accomplished more gradually over a gain range above (or around) the threshold.
The term hard-knee has come to mean only two discrete reduction levels: none and full ratio. Soft-knee on the other hand consists of: no compression, full compression, and a gradual curve to get from one to the other. Sometimes the knee is controllable with its own knob, and other times the compressor itself simply has a characteristic knee or a switch between hard and soft.
The way in which a sound ends can be just as distinctive as how it begins. Some end quickly, some slowly. Some give an amplitude swell, while others just fade away.
A control to adjust release time is often included on dynamics devices, to set the time it takes the compressor to stop reducing gain once the signal has fallen back below the threshold. As with attack time, in “auto” mode the manual release time is also usually bypassed.
Some compressors have a secondary input called a sidechain (also called the key, trigger, or control). When the sidechain input is activated, its signal is compared to the threshold and triggers compression of the audio signal (called the program material, can also be thought of as the carrier). The sidechain signal itself is not compressed, it just acts as a trigger.
This can be used to automatically balance two or more instruments relative to each other. When one performer comes in, the level of the other is automatically reduced. You often hear this on sports broadcasts, where the background noise is huge, and each time the announcer talks the background gets knocked down (or ducked) so that the announcer can be heard and so that the total broadcast volume remains relatively constant. Here, the background arena mics are the program material and the announcer’s voice is the sidechain (or trigger) signal.
When altered versions of the same signal are used for trigger and program, such useful effects as de-essers and pop reducers can be made. Again, the thing to remember here is that: one input triggers the compression but the other input is the one that gets compressed.
Output (makeup) gain
Output gain, often called makeup gain, is another control parameter common to compressors. This controls the level sent from the output of the compressor, after any compression has been accomplished. It is useful for boosting the signal back to a level appropriate for recording, mixing or reproduction. Remember compression increases the average relative level, but the overall level is brought down in the process and needs to be brought back up.
A bypass switch is the final function commonly found on compressors. While the basic role of this switch is to pass the audio from input to output without any compression, the specifics of its exact function can be deceiving. The biggest issue here is: what features of the compressor remain in the signal path?
If input and/or output amplifiers are still active, then both the level and tone of the bypassed output can differ from the input. Furthermore, in DAWs the latency (time delay) incurred by inserting the compressor may or may not be affected. Know how your gear functions in “bypass”!
Now, let’s take a look at metering before moving on to other kinds of dynamics processors. Compressors can have level displays for input, output, and gain reduction. Each of these may be expressed as either an average (over time) or peak (either instantaneous or over time). Average levels are useful for approximating the way we tend to hear audio. Peak metering is useful to determine the exact nature of what is happening and how it may affect the signal, especially in digital. (Metering was discussed in great detail in TCRM #5)
Knowing which type of metering your compressor uses will help you understand where your levels really are in relation to clipping and the next elements in your signal chain.
A limiter is a type of compressor that uses a ratio of 12:1 or higher. Some “brick-wall” limiters have a ratio of ∞:1, or infinity to 1! In this case, the limiter will not let anything pass the level of the threshold. They are handy if you want to get the hottest signal possible without exceeding the recording, mixing, or reproduction system’s maximum level.
Soft-limiting for digital audio
Soft-limiters are specifically intended to prevent audio which would exceed 0 dBFS from clipping and causing the associated distortion. Since this is a last-defense/safety type of function, these limiters generally have a very high threshold, often around -6 to -2 dBFS for digital applications. The ratio is ∞:1 and the attack and release times are both nearly instantaneous.
If a hard-knee limiter were used, it would simply flatten any part of a waveform exceeding the threshold and create the same clipping distortion it’s trying to avoid. For this reason a soft-knee function is used. This will keep the top of the waveform more rounded over a greater range of level above the threshold. Be aware, however, that when a soft-limiter is acting on the signal, it is modifying the waveform and can still cause noticeable distortion.
Because the threshold level must be exceeded before a compressor or limiter can start to react, traditional designs usually cannot act instantaneously. To make instantaneous reactions possible (or sometimes even fortune-telling) some designs now include a look-ahead function. This means that the processor is looking at a part of the waveform that will be coming through the compressor shortly to determine when to respond.
This is most commonly done by using a small buffer to delay the signal to be processed. An un-delayed version of the signal is used to trigger the compression function, but it is a slightly delayed version that is actually compressed. (Now if only there were a way to do this at the racetrack….)
Expanders and Gates
Though they have similar controls, expanders are really inverse compressors. An expander increases the dynamic range (rather than reducing it) by either reducing gain when the level drops below the threshold, or increasing gain when levels rise above it. The idea here is usually to reduce the noise floor relative to the audio signal. Expanders can have all of the same basic parameters as compressors. In fact, some compressors can also function as expanders.
A gate is a device that only passes signals that exceed a threshold. Audio that drops below this level will be silenced. Basically, a gate is an expander with a ratio of -∞:1. Also like expanders, gates are commonly used to automatically remove unwanted noise when a performer is not playing or singing. Their method of doing this, however, is a bit more extreme, and sometimes too obvious. Also like expanders, gates can have many of the same control features as compressors, but have no need of a ratio control. Many compressors can also function as gates.
TCRM 17 has outlined the features and functions of dynamics processors but has barley scratched the surface on how they can be used. I guess that will have to wait for TCRM 18….
John Shirley is a recording engineer, composer, programmer and producer. He’s also a Professor in the Sound Recording Technology program at the University of Massachusetts Lowell and chairman of their music department. You can check out his wacky electronic music CD, Sonic Ninjutsu, at http://cycling74.com/products/c74music/