This is a tutorial about the most overblown topic in music production. There’s hundreds of dynamics plugs out there in vst(etc) format, many more than any single other category of plug. Understandably the subject is a bit of a stumbling block for noobs.
This is a tutorial I wrote a while ago. It’s in two parts - part one is mostly for beginners, part two is mostly about advanced shit. So if you consider yourself halfway to being a hardcunt you may just want to jump in at part two.
The compression operation generally means “reducing the volume above a threshold”. So with this in mind it follows that it is just one of four similar operations to alter dynamic range. Occasionally you will find software that can do all of these operations.
The four operations are:
i) Compression - reducing the volume above the threshold. Extremely common, internet is ankle-deep in compressors.
ii) Expansion -reducing the volume below the threshold. This is a bit like gating the audio. The quiet parts of the signal are reduced, so you may use it to remove background noise, removing reverb, etc.
iii) Upwards Compression - raising the volume below the threshold. This is very uncommon. You could use this to raise the background sounds without destroying main transients.
iv) Upwards Expansion - raising the volume above the threshold. Sometimes compressors will run into “negative ratios” and function in this way.
Ok, let’s get back to standard compression, reducing above the threshold.
You can use compression for:
i) Making stuff louder, ie drum patterns and breakloops. If you squash the transients (the spiky bits of the signal) then you can turn the volume up for the entire signal without distorting. End result is a much louder piece of audio.
ii) “Glueing” tracks together. This brings the dynamic contours of several tracks together, making it seem a bit like the same sound. Allegedly this is pleasing to the ear.
iii) Taming a very long sound. As a hardcunt you may probably never do this. With very slow settings you can use the device to bring the volume down for extended periods (lasting many seconds or minutes).
iv) Surgical repair tasks on imperfect audio. You would also use other dynamic-range operations, ie expanding.
The most commonly seen variables on compressor plugs are going to be attack, release, threshold and ratio. These are likely going to be the main variables you use even on the most sophisticated compressors. You are generally also going to find a makeup knob otherwise you won’t be able compare the uncompressed signal to the compressed signal.
If you are a total beginner and don’t yet understand the attack / release envelope stuff, then I recommend you get hold of a compressor with a decent gain reduction graph - for example the fabfilter proC2. Play a raw drum loop at around 100bpm, change the attack and release, listen to the sound, look at the graph.
Why do we even need attack/release envelopes? The envelopes are necessary to “glue” tracks together. Also remember that in a truly live environment the processor can only respond after the sound is played. Finally, a zero attack, zero release setting will create distortion on many compressors.
Many compressors will have attack and release times printed in milliseconds. However the actual times you get will typically be quite different and depend on the internal modelling of the compressor. A 10ms attack on one plug might be equivalent to 100ms on another. Some compressors may have a % rather than ms. As a general rule use your ears and the display rather than the numbers on the dial.
GR stands for gain reduction. Most compressors will have some kind of gain reduction meter. Analog emulations may have very slow needles that are fairly useless when compressing a fast breakloop. The most recent software compressors tend to have fast GR graphs or meters.
The ratio knob tends to operate between 2:1 and 6:1. Above 6 it can be very hard to notice the difference.
To “glue” a few percussive tracks together you should let some of the transients get by. Make sure the GR returns to zero at least one place in each bar. When the GR doesn’t get back to zero for a while (ie several bars) then the compressor is essentially just turning down that stretch of audio by the same amount (and then turning it back up again). When a few tracks are compressed together they will have a common envelope and tend to sound more like one single piece of audio. A GR around 5db is usually enough to get a glue sound.
You should remember that any effect can potentially “glue” different sounds together. For example, applying the same saturation to several separate tracks will also create the feeling that they are stuck together.
When compressing a slower beat (ie 100bpm) you can poke and prod the knobs to make the sound “snappy” or “pumping”. If you are writing fast breaks (175bpm) compressor styles will be much less pronounced. A “fat’ sound usually refers to a high threshold, short attack/release setting.
The following is a list of advanced parameters you may find on various complicated compressor plugs:
* Knee - a soft knee of at least few decibels will usually sound less “artificial” than a harsh knee. You can probably just “set and forget” the knee. Rarely you may want to go harsh knee for some surgical editing or some such.
* Wet/dry - this allows you to scale the compression on and off. It allows you to do so-called “parallel compression” (I mention this again a bit below). The output gain is usually routed before the wet/dry. The internal graph/display on the compressor may alter when you turn the wet/dry or it may always show the max amount of GR.
* Range - this limits the amount of gain reduction and can potentially change the envelope shape considerably. On a fast drumloop the effect may be too subtle to notice. If you limit the range try to figure out how it relates to the attack and release as the implementation may vary.
* Peak hold - you may rarely discover this variable. Like range it can potentially change the envelope shape a lot but it’s probably not going to blow your mind.
* Channel separation / mid-side modes - This can create subtle special effects but hardcunts probably won’t bother with it. You could always route it inside the daw anyway if you think it’s worth doing.
* RMS/peak detection - RMS settings generally sound similar to a slow attack. Something wavery and acoustic might require RMS but if you are compressing drum loops you want to set it to peak.
* Automatic makeup - This is where the compressor automatically attempts to compensate for the change in volume as you move the main knobs. You might find it easier just to AB the sound.
* Automatic release times / program dependent times or ratios / feedback modes / fancy envelope shapes / fancy knee shapes - hardcunts won’t go here. If you for some reason think any of this is important then AB the sound.
* Lookahead - this one however, is good to know. Compressor plugs will often delay the outgoing audio by a tiny amount and report the delay to the daw, which compensates for the delay so it won’t exist in your track. The original non-delayed signal will be used to drive the compressor. One millisecond may be enough to get the jump on sudden transients. Occasionally you will find a variable lookahead. Having a big lookahead and a short release could really make some shitty sounds.
Running a lookahead of at least a millisecond is probably desirable when writing fast music. You may need a lookahead to properly squash a wack transient in a drum pattern. Or it may be desirable to have a long attack, long lookahead on some surgical-type alteration of a vocal sample.
You should be aware that (in my understanding at least) lookaheads on limiters are very different to the lookaheads you find on compressors. A limiter lookahead uses the time to scale the attack. Therefore a short lookahead on a limiter will be louder than a long lookahead, but with possibly more distortion.
* Oversampling - not compulsory but good to know. Oversampling will increase the capacity of the compressor to accurately catch the fastest sharpest transients. This sounds useful but mostly the difference is insignificant. When you can hear a difference it will only be at the top end - you may hear spikier hats for example, which might not be desirable anyway. In some cases I have seen bad implementation cause millisecond delays or bugs in the daw.
I covered the most common parameters you are likely to find. However there are squillions of compressors out there and there are a lot of idiosyncratic models with extra frills.
Some techniques used in single-band compression:
* Parallel compression. Also sometimes called New York compression. This is where you mix say half the dry signal in with a heavily compressed wet signal. The result may not be that different to just using the other knobs. Arguably the main advantage of parallel compression is that you can quickly crossfade between a heavily processed sound and a lightly processed sound.
Sometimes parallel compression is confusingly called “upwards compression”. When the wet signal is squashed very flat (with a high ratio and fast attack) then the lower volume parts of the signal are made very loud. Therefore if you mix this squashed/wet signal in with the dry you are boosting only the quieter parts of the dry signal. So it resembles upward compression.
* Highpassing the detector? You can EQ the signal used to trigger the compressor without EQing the actual audio. A common approach is to highpass the signal so that the compressor will respond less to the bassier sounds. In theory this means a kick drum will remain loud and punchy while other components are compressed. The kick however may have plenty of high frequency content and there are often cymbals etc sounding at the same time. The best you can hope for is a fairly subtle effect that only works on certain drumloops at certain tempos. Using an EQ on the detector may however be handy for doing some surgical correcting shit on rare occasions.
* Boosting transients? It’s not at all intuitive but by setting the attack and release precisely and with a high ratio and makeup you can actually increase the transients. This may be desirable in some circumstances. However you would get more pronounced results using an expander or a transient shaper.
* Sidechain compression. This is another technique you may hear about but as a hardcunt may not use. You can route an alternate signal into the compressor and possibly EQ it as well. The compressor will thereby respond to one piece of audio and compress the other. You can sort -of insert one piece of audio into another.
A common use of sidechain compression is ducking the bass in a house track to insert a big fat kick. It sounds much like a gater effect. Or you might duck the mids to insert some vocals. Essentially however this is “ducking” and therefore doesn’t belong in this tutorial.
Some general advice for compressor noobs:
*Always trust your own ears. Don’t use some compressor just because someone else recommends it. In my opinion, if you can’t hear the difference don’t bother.
* If you are doing something subtle (in dynamics but not only) you should use an AB technique. Switch between the clean sound (ie A) and the compressed sound (B) to judge the difference. Potentially you also want to CD and possibly EFGH. If your plug doesn’t natively AB you could possibly duplicate the compressor and switch between the two versions using hotkeys. You will have to set the gains so that A and B are at the same loudness.
* If you find yourself comparing compressor(etc) settings that are only slightly different you can do a blind test. Close your eyes and press the AB button several times so that you can’t remember or see which setting is which. Try and gauge only by your ears. Hopefully you only do this when you're at the mastering stage, otherwise you could just be wasting a buttload of time.
*To write breakcore the attack and release must be sufficiently FAST!!!! This means many software compressors are completely useless. You may want to avoid, for example, emulations of classic analog compressors.
* Don’t overdo it. If you are making a loop with already compressed drum samples you may be just overprocessing your sound by compressing them further. Similarly be wary about compressing things several times at different levels in the daw. You may get very horrible spiky attacks that you didn’t expect.
Is there any point compressing when the whole track is going to be squeezed down by a limiter/maximizer anyway? Well, yes. Firstly compression adds a release envelope which won’t entirely be destroyed by the limiter. Secondly the compressor is (probably) operating at a much lower ratio and with a much lower threshold than the limiter. The limiter is only going to cut off the top few db. Thirdly the punchy attacks of a compressed drum loop are still going to cut through other layers (ie bass, synths) before the layers are all limited together.
Compression unfortunately tends to reduce the low frequencies of a percussive loop. This becomes particularly evident when running fast compressors over breakloops. The sharp transients pass through and the compressor reduces the volume on the bassier parts of the drum hit. The best approach might be to pass the bass and only compress the tops, which is possible in the realm of multiband compression.
Serious nerd warning, read on at your own risk.
A multiband compressor splits the audio up into bands which are then processed in parallel. The bands are essentially split up by sharp highpass and lowpass filters. Each band gets its own compressor. There are very small regions of overlap between the bands, depending on the slope of the highpass and lowpass filters, but it will not be audible. The entire audio is put back together after the crossover and can potentially sound identical.
A dynamic equalizer is a parametric equalizer where the eq points are controlled by a dynamics process. The points could be a bell or a shelf. The dynamics will boost or cut the audio at the eq points. The audio is not processed in parallel and there can’t be any overlap between processes. It cannot process entire audio as seamlessly as a multiband compressor because there will always be points on the frequency spectrum between eq points which do not move. Even if the eq is divided into bands by using alternating highpass and lowpass shelves there will still be regions between the bands that will create bumps in the frequency response.
A mulitband compressor is therefore a better tool for “compressing” an entire drum loop or track. A dynamic equalizer is arguably a better tool for editing a narrow frequency range, for example expanding the high hats out of a mix. Also you can apply a “scaled” compression along a gentle eq shelf (for example).
Analog crossovers have one minor setback, which is phase colouration (code red nerd warning). You may have flipped the phase of a signal before to do something technical or to create an artificial stereo effect. The phase can be flipped anywhere between 0-180 degrees. Turns out analog filters create a phase shift every time you use them. The steeper the crossovers the more phase shift you get. Parts of the signal are added and parts are subtracted. At a low steepness of 12db you are struggling to hear any difference. At a high steepness ie 120db, the “phase distortion” is very evident. The audio now has a lot of colouration that it didn’t have previously. To sidestep this problem many crossovers will have a digital or linear mode. These crossovers do not create a phase shift, however they incur a latency issue. There will be a significant audio lag reported to the daw. In the mastering process this lag will not matter, during a live show it could be a problem.
Some crossovers have a “hybrid” mode which will (apparently) attempt to emulate the curves and “warmth” of analog crossovers without creating a phase shift. They will still involve latency (possibly negligible).
Some plugs operate with two compressors per band. The two compressors may have completely separate parameters or they may have linked attack and release times etc. With two downward compressors, you could gently compress most of the audio with a low ratio, and then limit the very loud parts with an infinite ratio (the Ozone multiband comp is set up like this). An alternate configuration would be to have below threshold dynamics and above threshold dynamics on each band (the native Ableton multiband is like this). Potentially you could apply both downward and upward compression simultaneously to each band which might help tame very erratic audio.
Where to place the bands in a multiband compressor? Theoretically you want to put the crossovers between areas which have similar amounts of dynamic activity. However as a hardcunt you may well have shit flying all over the place so just spacing them approximately is probably going to work fine.
Number of bands? As a general rule the more bands you have the “tighter” the sound is going to be. However beyond four bands it will be difficult to hear the difference.
Here’s some mulitband compression techniques that could apply to hardcunts:
* Squashing the top band can create a mashy cymbal sound reminiscent of the amen break (etc)
* With two bands you can separate the sub bass and use the top band to go for a snappy drum sound that cuts through other layers
* You could loosen the attack on one or more of the middle bands to keep it punchy in the mid range
* You can pass some (or all) of the lower band to keep a bassy sound (as previously implied)
* You can mash the sub layer down into a steady hum
* Whatever other pretentious crap you want to try
Ok, so what about limiters/maximizers?
You may think of a limiter as essentially just a compressor with a very fast attack/release, and an infinite ratio… however they typically employ very different architecture to compressors. Modern limiters use a lookahead to check the incoming audio for transients, and then use this period to scale the volume down very rapidly (in a millisecond or so). The algorithms vary from limiter to limiter and they are usually kept secret. The Waves L1 (released way back in the late 90s) was amongst the first plug of this kind, and since then limiters have become much faster and louder.
Basically speaking, the shorter the lookahead and the shorter the release, the louder the audio - because less of the signal is being suppressed unnecessarily. Transients may also seem snappier with shorter lookaheads. As the times get shorter and the limiter gets faster it approaches hardclipping (which may not be too bad in some circumstances). Generally breakloops typically sound good with a short lookahead (hardclipping or thereabouts) and a more forgiving release stage to avoid distortion.
If you have the limiter set decently, you can usually just push the audio into it and make it instantly loud. However, like compression, heavy limiting will tend to reduce the bass, and when pushed hard the top end gets a little crunchy. Some artists seem to purposely go for this sound.
To hear exactly what the limiter is doing you can turn down the output at the same time as you raise the input (link the controls or turn the knobs at the same rate). If you are a real hardcunt you can make the audio as loud as it can possibly be by doing this and listening for the point where the audio begins to sound quieter. Then raise the output again.
Modern limiter design varies enormously from model to model. You will have a threshold/input control but that’s about all you can guarantee. There may be controls for the lookahead and release, or you may have similar controls unique to the design. There may be loads of parameters or only one. You will also come across multiband limiters (for fuck’s sake).
Intersample Peaks? (Nerds only paragraph). Heavy limiting (apparently) inadvertently produces intersample peaks. If the digital audio is recreated by an analog amplifier, it may have values above 0db. This can (allegedly) create clipping/distortion. If you use a true peak monitor you can see the intersample peaks and if you oversample the limiter you can reduce them to a fraction of a decibel. How important this is I do not know, but some limiters have a true peak mode so I thought it could be worth mentioning.
And what the fuck are transient shapers?
Transient shapers analyse the incoming audio and decide what parts are transients. Typically they use a very short window (ie 40ms) and look for the loudest parts. The transients are then typically labelled as “attack” and the inbetween areas are often labelled “sustain” (this is obviously confusing terminology if you have just spent a few hours learning compressors). There may be a decibel value for the attack and sustain stage. Thereby a transient shaper can sound like both an expander and a compressor. Depending on the secret algorithms these plugs can sound amazing or subtle or sometimes cheap.
There may be a variable for the window time which obviously you want to set roughly to how fast the transients are coming. A big window will catch only the biggest transients (maybe the snares and kicks) while a small window will catch the inbetween transients too (maybe the hats and percussion).
Other transient shapers (ie Ohmicide, Saturn) make things simple and have essentially just one knob that goes between fat and thin.
And what is loudness?
There is a lot written about this elsewhere so I’m only going to quickly mention it here. Obviously by compressing and maximizing a piece of audio it becomes louder, even though the volume in decibels remains the same. So looking at the faders in your daw is not a good indication of how loud your track is. There seems to be a few competing algorithms out there to measure loudness, but as far as I can tell the best is a standard generally known as LUFS. Short term LUFS means the reading is averaged every three seconds. Many plugs are available that show a reading on this scale, go to the options tab of Ozone for example.
A contemporary electronic track will typiucally have a short term LUFS that hangs around -8db. Extremely loud tracks are up at -3db (this is the maximum anything can go but some tracks are released this loud). Tracks with a lot of dynamic space are lower, ie -15db.
1 post • Page 1 of 1