The inverse of compression is expansion.
Compression is used a great deal more often than expansion, but the expansion technique is very helpful in certain situations. I mentioned this in yesterday’s blogpost briefly, but here’s the full deal on how expansion works.
Picture yourself in a room.
The ceiling represents the very loudest peaks in the music, and the floor represents the very quietest moments.
No processing is applied, and the floor and ceiling are where they are.
You see the ceiling is the highest point in the room and you see the floor is at the lowest point.
The difference is the height of the room, representing the dynamic range in this metaphor.
So, how does the room’s height (dynamic range) change if you use compression or expansion?
Compression makes the ceiling start descending towards you, coming down on top of you
Expansion makes the floor act like a descending elevator, taking it further away from the ceiling
Compression reduces dynamic range by turning down the loudest bits.
Expansion increases dynamic range by turning down the quietest bits
Or, in other words,
Compression reduces the distance from the peaks to the quietest moments.
Expansion increases the distance from the peaks to the quietest moments
Just remember that with compression you turn down only the loud bits, and with expansion you turn down only the quiet bits.
It is the dynamic range present in the signal that is being compressed or expanded.
There is an expansion ratio control, just like the compression ratio control, so that you can decide the rational quantity of change.the ratio between input level and output level once the effect is triggered. You will find ratio on both types of processor, then.
As I said yesterday, expansion ratios are always lower than 1 to 1 (e.g. 0.5:1) and compression ratios are always higher than 1 to 1 (e.g. 3:1).
Expanders are handy for reducing annoying headphone spill at the end of vocal lines, but it can be tricky to do. Keep the loud bits (the singing) and reduce the level of the quietest bits (the spill from headphones between vocal lines).
The threshold must be set carefully for this to work well. You want to keep the ends of vocal lines, but you want the spill that is most noticeable gone, and it is really noticeable just after the singing stops. Set the threshold so that you can keep the singing and lose the spill.
The same technique can reduce unwanted reverb a bit by attacking the tails of the sounds. It’s not always successful.
Many expanders have sidechain or key input features, just as many compressors do. More in a moment on that very helpful feature.
A noise gate is a processor that acts automatically to silence any signal that falls below a predetermined threshold level.
Sometimes, you may find a control that lets you set a predetermined level reduction rather than silencing the sound completely.
It’s called a gate because it was first designed to simply open or close, allowing signal to pass or not based on how quiet the signal was at any moment.
If the signal is above the threshold level you’ve set, then the gate is open and signal will pass through the gate and continue on it’s merry way. Once the signal falls below the threshold the gate closes, or causes a preselected amount of gain reduction.
Turning something way down is a lot less disturbing to the human ear, so it is a preferable technique in many cases.
Gating draws attention. This is it’s strength and it’s weakness. A popular use for gating is to cut off the reverb on individual drums, as in the classic Phil Collins sound from the 1980’s. This is called gated reverb, unsurprisingly.
At the top of a mix, there may be an open microphone or three on a guitar amp that buzzes and hums. These sorts of noises are cumulative, and the more you have the worse the overall sonic quality gets.
You may or may not want this at the top of your song – or at the end in the fading last moments. Noise gates to the rescue!
Using a DAW to edit the silences in yourself where you want them is preferable in both these tasks, being precise and easy to do, but in the olden days we used to use the noise gate for this purpose. You still can if you like. It’s another tool.
It’s very useful in multi-miked drum recordings to minimize spill between different microphones on different parts of the drum kit. Selecting, placing and aiming the microphones well is a very large part of this too, but gates are another useful tool in the recording toolbox to tighten up a drum kit recording.
This automated silencing control can be a very useful thing, but it has a musical downside, since the sudden silencing of a signal will draw unwanted attention. It can feel very abrupt, which may be unmusical in context or undermine the emotional intent.
A noise gate can be triggered, like many other dynamics processors, by listening to (and responding to) a signal present in a separate input path called a sidechain. Depending on what that sidechain signal does, the processor treats the main signal accordingly, rather than the sidechain signal. I’ll explain this sidechain thing in full in a moment.
NOISE GATE CONTROLS
You’ve already met the common controls on compressors. Here are controls that appear on noise gates alongside those also found on compressors (like threshold, att and output level).
Hysteresis and range are hold are controls found on many noise gates.
Range sets the amount of reduction that will occur. Sometimes this is very small, resulting in a small drop in level, and sometimes it is very large, and may be large enough that it results in total silence at the output. With range you choose how low in level you want to make things get.
Hold is a control allowing you to hold the gate open for a chosen amount of time after the closing of the gate has been triggered.
Hysteresis is a nice word that belongs in a lyric one day.
If you mention hysteresis, see the look on most folks’ faces
What is it?
Well, noise gates are opened according to their attack time parameter, and closed according to their release time parameter.
There is also a hold parameter so you can have the gate wait a while before the release phase begins, which can be very helpful in avoiding false triggering of the gate if the levels are hovering around the threshold level, regularly crossing it.
This repeated crossing of the threshold can cause an effect called chattering, where the gate stutters on and off repeatedly as the signal level goes back and forth across the threshold.
Hysteresis is a control to combat this chattering. It’s usually a better way to solve the problem than fiddling with attack and decay controls. It lets you set a second threshold level. The first triggers opening the gate, the second triggers closing it.
Hysteresis is the difference between these two thresholds, meaning the range of values between the two levels. You can widen or shrink this two-threshold window to combat the chattering by raising or lowering the hysteresis parameter.
Not too many of these around, but the Sonalksis SV-719 Expander/Gate/Ducker is one of the best. It is a plug-in that can be set to do ducking, and there’s a nice screenshot of it at the top of this post.
Ducking is automated reduction of one signal level (MAIN SIGNAL) in response to the properties of a different signal, the sidechain input signal (KEY INPUT).
The ducking of the Main Signal is followed by rapid return at a predetermined rate to the level required when the Key Input is not present (more accurately, below the user-set threshold level).
The classic example would be turning down the music whenever the announcer speaks over it, so that you hear the voice at all times over the music. The music jumps up rapidly to the normal level again during any sufficiently long pauses in the vocal signal.
You hear this on TV and radio all the time, especially during advertising commercials. It’s an intrusive effect in this context, and deliberately so. In record production, it is usually a lot more subtly applied. You can use it to have the snare and/or guitars reduce in level a bit if the singer is singing at the same time, for example.
Maybe a 3dB duck in the fingerpicked acoustic guitar during the vocal lines will help. And so on.
So here’s a bit more detail on what makes techniques like automatic ducking and de-essing possible.
SIDECHAINS AND KEY INPUTS
The sidechain feature of a dynamics processor is invaluable. It’s not actually a control, but it does make a huge difference to what happens.
You won’t find it on all dynamics processors, but it is extremely useful. Many dynamics plug-ins have a sidechain ability, since it’s trivial to route or duplicate signals from place to place around a DAW environment.
In normal use, the compressor reacts to the input signal crossing the threshold level.
There is an additional input to the compressor if it has a sidechain, called the sidechain input, or, more commonly, the Key Input.
By feeding a signal into the key input, the processor will respond instead to the Key Input signal’s level, and whether THAT signal crosses the threshold you have set.
Let’s call the main signal you are trying to compress SIGNAL, and the Key Input signal we’ll call KEY.
The most important thing to understand about sidechains (key inputs) is that the compressor is listening to the key input signal, but it is not acting upon it. Instead, the compressor responds to the key signal and processes the main signal depending on the properties of the key input. It “hears” the KEY do something, then treats the SIGNAL as if it had done what the key did.
This may seem puzzling, but actually it’s pretty straightforward.
Put an EQ filter into the sidechain so that the key signal can be optimized for triggering the effect. You can’t hear the key input in the mix, although there will be a Key Listen button or mode that lets you hear the key signal with the EQ or other processing applied to it.
TIGHTENING ACOUSTIC KICK AND BASS GUITAR TRACKS
For example, insert a noise gate plug-in on a bass guitar track. Choose the kick drum track from the list of other tracks shown in the drop-down menu of your key input selector (in the noise gate plug-in on the bass track). Select the low-pass filter and high-pass filters in the plug-in so that the LPF and the HPF are applied to the key input. Use Key Listen to set the highest frequency you will want to allow through the sidechain.
With a kick triggering the gate on a bass track, you could throw away everything above, say, 800Hz and everything below about 40Hz. That gives you plenty of the kick energy but it removes prominent snare drum spill that may be on the track. Have the fastest attack you can on the gate.
Now you can have the gate stay closed until the kick drum triggers, using appropriate settings, and at that instant the bass will be allowed through, letting you have a kick and bass perfectly timed to start in sync on each kick beat, rather than one anticipating the other.
There are dynamics processors that operate only when a user-specified frequency range is louder than the chosen input threshold level. The “de-esser” is the typical type.
You do have to tune them to the frequency of the sibilants of the particular vocal take, which is not so simple to do, even for experienced engineers as it is often somewhat hit and miss to find the correct frequency area to clamp down on.
The de-esser is a clever application of compression that uses an equalizer in a sidchain much as with the kick and drum example above.
It is a “frequency-conscious” compression method. Compression only occurs at a specified frequency or range of frequencies is the frequency-conscious method.
Using a spectrum analyzer can help a lot with finding the frequency you need to turn down. Run the sound through one of those and look for the energy peak on the “esses” or “effs” somewhere between 2kHz and 8kHz (most commonly around 5kHz or 6kHz).
Find the right one, and then tell the plug-in (using the sidechain) what it is so it will turn it down whenever it senses too much energy there during the song.
Hopefully, only the “esses“ will be affected by this, but its not exactly targeting “esses” as such, just the abundant energy they have in a certain area.
Careful setting is pretty effective at getting decent results, but, in practice, that frequency will be affected whenever the de-esser’s input threshold is passed, which need not be only on the “esses”, of course.
You can get much more reliable (and less intrusive) results by specifically editing only the “esses” in the vocal track, one by one, using fader automation to momentarily turn them down, then back up for the rest of the word.
Manual gating is popular, a term for editing the silences into your DAW track one by one, rather than using a plug-in on the track.
Again, the reason for not doing this would be it may be very time-consuming, depending on the part. Typically, the DAW also has Strip Silence features which offer parameters you can set to finesse this action, based on audio levels.
It’s also rather abrupt when you simply chop parts off to silence. Here, your DAW has fade-ins and fade-outs to help you.
Apply a rapid gentle fade-out, shaping the fade curve to taste, instead of using an abrupt end.
USING FADES TO MIMIC GATING ACTION
Check your Edit menus with the take’s region or file selected in the Edit window/page of your DAW software. There should be a Fade… option which will allow you to select various fade-in and fade-out shapes.
You can usually choose between different types of fade curve (a graph of desired level behaviour over time), even if they are only switchable between linear or fixed concave or fixed convex. Experiment and see which types sound best to you in different circumstances. There may be simply a numerical value or an onscreen Fade knob to “turn”. Apply to taste.
CAUTION: USE AT LEAST 15ms CROSSFADES, AND DON’T EXCEED -0.5dBFS LEVELS
When you splice two regions together with an edit in a DAW, you should always be careful to make a 15ms crossfade between the two to prevent the introduction of spurious clicks and pops at the edit points.
This problem is due in most cases to an abrupt change in level at the individual sample level at the edit point, particularly if one is a downward-descending waveform and the other region is moving in an opposite direction away from the centre line.
When an edit to a digital audio file causes a sudden shift from the positive to the negative side of the centre line of the waveform display, or vice versa, clicks or pops are bound to arise.
Simply having a different level either side of the join causes this issue too, even if the waves are trending in the same direction either side of the join.
Putting a crossfade across the edit point is the solution, and anything shorter than a 16ms crossfade will be at risk of audible problems.
Finally, I’ve found that it’s surprisingly easy to miss occasional instantaneous digital clicks and pops, only noticing them when the song has been on sale for a week.
Their presence also limits the maximum level that can be obtained in mastering, since the file appears to have 0dBFS Full Scale peak transients at the pops and clicks. This in turn can lead a lot of playback equipment (mp3 players and the like) to create intersample peaks that are actually momentary distortion rather than audio.
The file may not distort, but it’s converted back to analog by playback equipment to be playable on analog speakers. At this time, the equipment has to fill in the gaps between the discrete samples of the file and make a continuous analog version. The filling is the interpolated idea the equipment has of what must be between the two adjacent sample levels at the edit.
This is where you can get even more pops and clicks than you had to begin with. If two adjacent samples are trending up above the 0dBFS mark and are themselves at -0.1dBFS or higher in level, then trouble can arise when the playback equipment runs the mathematical algorithm it uses to figure the levels out at conversion to analog.
Phew! Stick to levels below -0.5dBFS (that’s half a dB) with all signals all the time and you will be fine.
However, if you listen three or four times through any edit point carefully, paying full attention for clicks and pops, you will be
Be very careful to audition any crossfade carefully. If your crossfade time is 15ms or greater, then you should be fine.
Headphones are the most successful tool for making out pops and clicks that shouldn’t be there, but don’t rely on headphones alone. If you’re using only your monitors speakers, listening intently at low to moderate levels is also very helpful for detecting details, though headphones certainly win this one.
The idea is to run a firmly compressed version of a track alongside an unprocessed version, blending the two together to taste. This is the so-called “New York” compression method, nowadays called the parallel compression technique.
Duplicate a track in the DAW, then insert a compressor on the duplicate.
Lower the level of the duplicate to somewhere around 10dB lower than the original, and make sure that any level changes automated on the one track are in place on the duplicate.
If you duplicated the track in your DAW with automation data already written on the original track, it will copy the automation along with the audio regions onto the new duplicate track by default. If that does not work, check your preferences or options menu to see if the behaviour is turned off.
This technique really fattens up basses, drums, vocals, pretty much anything. Don’t get too loud or too quiet with the duplicate. The idea is to support the original and there is a sense of fattening up it’s sound when the two are combined.
It will sound beefier, thicker, fatter, more solid – you get the idea. Drums and vocals particularly like this method.
One last point here. It’s critical that the two tracks are in phase. They cannot be misaligned on the timeline and they must play back with delay compensation features switched on so there is no latency between them.
Low frequencies will be weaker, not stronger, if this is overlooked. They must be in phase with each other.
If you just duplicate the track, it will be in phase by default, but adding the compressor to only one of the tracks might require you to add another instance of the same compressor across your original track, but with this one set to bypass mode so it does not affect the original signal.
This should match the latencies of the two tracks, i.e. ensure it takes the same time for each track to process it’s audio if you have no delay compensation feature available and the compressor plug-in has any latency. Check in mono on a small speaker to be sure all is well.
If your mix has any phase problems, parallel compression of low-frequency content is one of the first things to check for discrepancies in latency or phase.