PLEASE NOTE: This article has been archived. It first appeared on ProRec.com in August 1998, contributed by then Editor-in-Chief Rip Rowan. We will not be making any updates to the article. Please visit the home page for our latest content. Thank you!
Limiters, gates, expanders, multiband compressors, de-essers, normalizers, ultramaximizers. These are all variants of the basic compressor, and they all can do a lot to help you get the most out of your mix.
If you missed part one of this series, go back and give it a read. It will help explain the basics of compression – what it is, why it’s used, what the controls do, and when to use it.
In this issue, we’re going to look at the rest of the dynamics controllers and see what they can do for you. And we’ll get a little deeper into our understanding of what dynamics processors do.
To better explain these tools, I’d like to introduce the concept of transfer function. A processor’s transfer function can be thought of as a mapping of input levels to corresponding output levels. If I put in a -12 db signal, when comes out the other side?
Here’s a graph that shows a transfer function for a compressor in “bypass” mode. Input levels are across the bottom, and output level is shown by the numbers on the left column. You can see there is a 1:1 correspondence, because the compressor is turned off.
Here’s a transfer function for a 3:1 compressor with a threshold of -20 db. As you can see, levels under -20 db are not altered. As the signal passes -20 db, it is attenuated by a factor of 3:1. You can also see the 5db makeup by looking at the bottom left. The line does not meet at (-100, -100) but rather is at (-95, -100) – meaning, that the entire signal has been boosted 5 db.
Now let’s look at these other dynamics processors. Let’s start with limiters.
The limiter is a kind of compressor. The purpose of the limiter is to provide control over the hottest peaks in the signal. The limiter is used primarily to get the most sound possible onto the media – whether the media is analog tape, FM, or digital. Moreover, unlike a compressor, the goal of the limiter is to change the dynamics as little as possible. The sound should still feel the same.
Limiters typically have only two controls: input, and threshold. The input controls the signal level going into the limiter. The threshold establishes the level above which the limiter will attenuate the volume. Sometimes a limiter will also have a release control.
Here’s the transfer function for a limiter:
As you can see it looks a lot like a compressor. A limiter can be thought of as a compressor set up with a high ratio (10:1 or greater) and fast attack and release (<10 ms). Notice how the transfer function becomes flat above -10 db. This means that no matter how much louder the input gets, the output will not rise above the threshold. The main difference between the limiter and a “hard” compressor is the attack and release time. The fast attack and release of a limiter means that volume reduction begins as soon as the signal crosses the threshold.
Used properly, the limiter will cut off the loudest peaks in the signal without changing the dynamics of the rest of the audio.
Here’s a drum loop that I had handy. Before, and after limiting.
Before limiting, the peaks of the first and last beat were at -1 db. After limiting, the peaks are at -6 db. The sound is identical to the ear.
When we record to analog tape, limiting is usually not an issue. That’s because analog tape functions as a limiter. Excess signal simply gets softly attentuated by the tape, because the tape can only handle so much magnetic saturation. So when rock bands record to analog tape, every track is limited.
External limiters are very useful when recording audio to digital media. The ability to “run hot” without having to worry about digital clipping allows the engineer to gain 5 to 10 dbs on each track. That adds up to a hotter, clearer mix. Put the limiter in the signal path between the mic preamp and the soundcard / converter. Then turn up the preamp to make full use of the available dynamic range.
Limiting is also often performed on the final mix. You can use a limiter to add a significant amount of volume to the final mix. This function should be performed by the mastering engineer, not the mixing engineer! Limiting adds volume to the music in the same way as any compressor – by lowering the peaks, the overall track can be boosted. Mastering engineers will perform compression first, if the track is generally too dynamic – then perform limiting last, to get the track up to maximum loudness.
Digital tools have created a new class of limiter: the zero-overshoot “lookahead” limiter. Whereas a standard limiter works like a compressor – by reacting to the increase in volume during musical peaks – a lookahead limiter anticipates peaks by analyzing the audio, and attenuating the volume just as the peak occurs. This kind of limiter can produce the heaviest limiting with the fewest artifacts. This kind of limiter is sometimes called an “ultramaximizer” – whatever the name, it’s still a limiter.
If you’re using a limiter while you’re cutting tracks, you don’t want to over-limit. A reasonable amount of limiting is about 3-6 dbs. This is sufficient to more than double the amplitude of the track without getting nasty. Remember that a limiter works best on dynamic program material. Don’t use a limiter on accoustic guitar or violins. On electric bass guitar, a limiter can reduce the “thump” of the string being hit, getting more of the note instead. On drums – especially snare drum – a limiter will reduce the “crack”, allowing you to get more of the tone of the drum.
Limiting the final mix should be performed by the mastering engineer. What if you’re the mastering engineer? Maybe you have a demo, that isn’t going to be mastered yet, but that needs to be “hot” and loud. How much limiting do you use? My rule of thumb is to limit harder and harder until you sense the drums losing power. When a mix is overlimited the drums will just start to vanish. Back off until the punch of the drums returns. That’s the right amount of limiting.
Multiband / band-specific compressors are another kind of compressor. This kind of compressor compresses audio within certain frequencies. The difference between a band-specific compressor and an equalizer is that an equalizer attenuates all instances of the specified frequency, while a multiband compressor attenuates the specified frequency only when the threshold is crossed. Some standard compressors have a sidechain. You place an equalizer in the sidechain to select the frequency to compress.
A de-esser is a band-specific compressor which is designed specifically for the sibilant frequencies produced by human speech. De-essers react to frequencies in the range of 2 Khz – 16 Khz. The range around 3-6 KHz is the most important, because the ear is most sensitive to those frequencies, and they can sound really nasty when overamplified. Because it is a compressor – not an equalizer – the de-esser only reacts to excessive treble. The de-esser can really help your vocals. When used with a good EQ, you can boost the overall treble to add breath and sparkle, while controlling excessive treble only on the consonants that become shrill. The overall effect is a wetter, shinier sound that never gets harsh.
A gate is a reverse compressor. Rather than attenuating loud signals, the gate attenuates only the soft signals. It’s often called a noise gate because it is usually used to eliminate noise when the desired signal is not present.
The transfer function of a gate shows that as the signal level drops below the threshold, the signal is muted. Gates are good to eliminate spurious noise in the track. For drum miking, the engineer will often use a pair of ungated overheads, to capture the sound of the whole kit. Then each drum is individually miked and gated and recorded to seperate tracks. Each track will be gated so that only the miked drum appears in the track – the other drums are gated out. Want a little more floor tom? You now have a track that contains only floor tom – turn it up a little.
There are other uses for a gate. Phil Collins, INXS and other artists in the 80s made famous a sound of a gated reverb applied to drums. Crank up the reverb, then gate the reverb’s output. You’ll hear that classic sound from “In the Air Tonight.”
Gated instruments can get wierd. Because the gate cannot tell the difference between good and bad low-level signals, the decay of an instrument – or a slow attack – will get muted. Some instruments cannot be gated, because they are too dynamic, and produce desired low-level signal.
Which brings us to expanders. A downward expander is like a gate that – instead of muting the signal as it passes below the threshold, it attenuates the signal. An upward expander is like a reverse compressor – as the signal passes above the threshold, the signal is amplified.
The two processes are similar, yet important differences exist. As you can see, the downward expander leaves the “audible” portion of the signal untouched. This is best to reduce noise on signals that cannot effectively be gated. The upward expander enhances the dynamics of the loudest signals, and can be used to add punch or attack in a “mushy” track. Downward expanders are often found as features on gate units, and upward expanders are often found on compressor units.
The last item on our list is the normalizer. Normalizers are found on every digital workstation. The standard normalizer is NOT a dynamics processor. A standard normalizer is only a volume control to turn the volume of the track up as far as possible without clipping.
Common practice is to never use normalizers. A normalizer is a destructive process that really does nothing constructive to the track. If the track needs to be louder, turn up the volume on the software’s mixer.
Some software, however, incorporates compression into its normalizer functionality – so I thought it would be valuable to cover that ground. Think of a compressor / normalizer as a standard compressor that has an automatic makeup control. Instead of adding makeup manually, the normalizer adds whatever makeup is required to bring the peaks up to 0db. I think this is a great feature that all software compressors should have. But, really, normalizing should be a compressor feature. Compression shouldn’t be a normalizing feature. This author writes: get rid of the normalizer!
Well, I hope we’re all still learning. I know I am. Next month we’ll review a handful of software compressors: Sound Forge, Cakewalk, Cool Edit, Waves, Gadget Labs and Hyperprism, and see how each one compares to the others.