Breaks, Old Skool, Nu Skool. Interested in how it's all done or have advice to offer? Post here
User avatar
Oldskool Raver
Oldskool Raver
Posts: 665
Joined: Fri Nov 11, 2005 9:20 pm


Post by DIEMOS » Sun Aug 17, 2008 12:49 am


Compression 101 : The Basics

One of the most important things to learn while learning how to use a compressor is: learning when to use a compressor. If you're writing a song, every element in your song will not always need compression. You can use a compressor to make things "fat", you can use a compressor to make some things "thin", you can use a compressor for "color", you can use a compressor to just trim off some unwanted peaks, and some people use a compressor as an effect to make the music "pump". There are many uses for a compressor, I'm going to attempt to show you some basics here today, and I'll leave it up to you to find new ways of using a compressor. Lets begin with a question:

What does a Compressor do?

A Compressor can be used for a few different purposes, one of the most common is to turn loud sections in your audio lower without affecting the already quieter passages. For example, when your audio has very large volume peaks but the average volume of the sound, the RMS, is not very loud, you could then apply some compression to bring the peaks down and then boost the overall volume of it to increase the average volume.

Here is an image I made of some audio where the first half of the clip was so much louder than the second half, that I could not turn it any louder because the first half would then be way too loud and clip. After compressing the audio though, you'll see that only the audio that was above the Threshold was lowered in volume, and the whole thing is a bit more leveled out now. I can now boost the overall signal louder without worrying about the beginning going over 0dB and clipping so quickly. (dont worry ill explain the Threshold next)


Ok, you see that line on that image above labeled "Threshold" ?

Explaining "Threshold", "Ratio" & "Gain":

Imagine you say to the compressor: " Set a mark at -15dB and if any audio goes over that mark, I want you to lower the volume according to how much volume reduction I tell you take off in the Ratio, but dont lower the volume right away; Take a little bit of time to fully Attack the audio where you'll go from no compression to full compression in a certain amount of time I tell you to, like 200ms for example. So once some audio has gone over the -15dB mark and you've taken 200ms to fully start compressing that audio, I want you to keep compressing the audio until the original volume goes below the -15dB mark again. Once that happens, I want you to Release the audio from your compression kung-fu grip (lol) and bring it back to normal volume, but dont Release it right away... take about 200ms to gradually go from full compression back up to no compression again. "

Ok, haha, let me get a little more specific now. That -15dB mark that you told the compressor to set is called the "Threshold", any audio that goes over it will be reduced in volume according to the Ratio, any audio that goes below it will pass through unaltered. When audio goes over the Threshold it triggers the compressor to begin working; if no audio ever goes over the Threshold, no compression will ever take place.

The Ratio is where you define how much gain reduction, or in other words volume reduction, you want the compressor to apply to the audio once the compressor has been triggered by audio going over the Threshold. The Ratio has two numbers, you can think of it as the first number being the input amount of decibels and the second number being the output amount of decibels. The second number (the output) will always be a 1, for example 4:1 is a Ratio and it's basically saying that for every 4 decibels that go over the Threshold, only 1 dB will be output. So, for example, if the signal triggers the Threshold by going 4 decibels above it and the Ratio is set to 4:1, the output signal will only be 1 decibel above the Threshold instead of 4. If instead the signal were to increase by 8dB above the Threshold, the output will be 2dB above the Threshold with that same 4:1 Ratio, and so on. (8dB divided by the Ratio 4).

After lowering any loud passages in the audio by compressing them, you can then use Make-Up Gain, which is simply volume that you add to the entire signal to sort-of "Make Up" for all the now-lowered peaks that you lost from compression. It's just volume that you can choose to add at the end of compression to bring the audio level back up, just incase you lost alot of volume due to compression.

Lets talk more about "Attack & Release"

The Attack is basically the amount of time that you want the compressor to take, in milliseconds (ms), to go from no compression to full compression once the signal has gone over the Threshold. During this Attack time, the compressor will use the milliseconds you define to increase compression little by little within that period of time until it reaches a "full compression" stage at the end of the Attack time. The Attack stage compression lasts until the audio drops below the Threshold again. Once the Attack stage has ended and the signal has dropped below the Threshold, the Release or "recovery" stage begins.

The Release function is a special one. While the Attack function compresses only the audio above the Threshold, the Release function is a bit different, it actually compresses the audio that falls below the Threshold. During the Release stage, the compressor will automatically detect where the audio is when it falls below the Threshold and begin to compress a little bit of the audio there. The Release function that you adjust on a compressor is the amount of time you want the compressor to take, in milliseconds or seconds, to go from that little bit of compression back up to the original, uncompressed, audio level. If any peaks go over the Threshold while the Release time is still active, then those peaks will be compressed until the Release time is over.

Sometimes a little bit of compression right when the audio drops below the Threshold can help the compression sound a bit smoother, some might say "natural" or "musical". Also this bit of compression can help mask any hiss or noise that appears when the signal suddenly drops below the Threshold.

The Release stage can serve many purposes, just to name a few:

You can use the Release stage to control (compress) certain peaks that follow the Attack stage once the signal drops below the Threshold. How long it lasts is dependant on your Release Time.

The Release can be used to accentuate certain things. When some material is highly compressed and the audio is sounding a bit flat, that little bit of compression in the Release stage can actually build some movement into the volume.

With long Release times on certain audio material, you can somewhat build a bit of dynamics in some audio by gradually going from compressed to uncompressed in the Release stage instead of having a flat section of audio.

If you dont want that bit of compression during the Release stage, you can always use the fastest amount of Release time possible on your compressor. I will note that, by using a combination of both a fast Attack and a fast Release (usually both under 50ms) a compressor can create distortion, because it then attempts to follow the actual waveform of the signal instead of the general shape of the audio.

Have a look at this audio block I designed, this block of audio is uncompressed:


Alright, now have a look at this one:


In this picture above, you can visualize the type of curves applied by the "Attack" and "Release" functions over the amount of time you define.

Here, in Figure1, you can see a section of audio that is about 200ms long with an Attack of about 100ms in time, notice that the Attack time curve ends at just about half-way through this 200ms section of audio at the beginning and thus has a steeper curve than the other 2 images.

In Figure 2, you can see an Attack time of 200ms and notice how the Attack curve is using that whole little 200ms section of audio to go from the original volume to full compression at the end of the curve.

In Figure 3, you can see the Attack time set to 400ms, but the section of audio is only 200ms in length? Well then the compressor wont have enough time to reach full compression within that small 200ms section of audio and will only reach about half way through gain reduction before the compressor detects a drop below the Threshold and kicks into "Release" or "Recovery" stage.

The Release stage, you can see clearly here in all three images. After the signal drops below the Threshold it will still compress a little bit, how long, depending on the amount of time you specify, in milliseconds or seconds, to go from that little bit of compression back up to the original, uncompressed audio level.

"Attackin" The "Action Snare"

Earlier you saw that I was able to level out a section of audio by bringing down the peaks in the beginning half of an audio section. In that situation I used an extremely fast Attack to start compressing the audio at the first sign of any audio going over the Threshold. But, there will be times when you're compressing something and you might want to let a little bit of the original audio get through before being fully compressed. In a situation like this, you can set a not-too-fast Attack time, all depending on the audio, to make sure some of the original, non-compressed, audio gets through before being fully compressed.

For example, say I was compressing some drums like the infamous "Action Snare", this snare is a bit different from alot of other snares, the "Action Snare" has a full body of noise that follows the attack of the drum hit. This particular snare has a very snappy attack on the drum hit, but because its followed by a very loud noisy body, the snappy attack is somewhat overshadowed by this full body of noise that follows it. In this situation, if you want to bring out the snap in a drum hit like this you could set a short/medium length Attack time, not too fast because you want to let some of the original audio through, but not too slow because you actually want to compress the body of noise to accentuate the snap in the attack of the hit. So here's what I decided on for this snare hit:

Threshold: -30db (A fairly low Threshold to make sure that the audio remains in the Attack stage long enough to compress the full hit and to make sure it doesnt go into the Release stage too quickly by dropping below the Threshold. So with a low Threshold like this, the drum hit will have to be almost finished before it drops below the Threshold and goes into the Release stage)

Ratio: 4:1 (again the ratio all depends on how much of the signal is going above the Threshold, right here 4:1 is all I really needed before I started to hear a difference in how much compression was taking place.)

Attack: 220ms (I settled on this value after sliding the Attack knob back and forth to hear at which point it started getting a bit more snappy and less noisy, I was also moving the Threshold up and down a bit to fine-tune the sound at this point.)

Release: 100ms (The Release time was not very important in this situation, but i set it to a safe 100ms to avoid the Release time still being active before the next hit of the snare comes around. I wouldnt want the Release time being so long that by the time the next snare hit comes around, it's still compressing and squashes my snappy Attack.)


Please note that, in this situation, most snares' Attack times are usually much shorter--around 30-50 ms to let the attack of the hit through on a common snare--but this snare is a very unique one with a long body of noise that required a very long Attack time for the compression to sound natural. I could have used a higher Threshold and a shorter Attack time but then you would hear the initial Snap followed by a "brick wall" body. So instead, with a longer Attack and lower Threshold, I utilized the shape of the Attack curve I showed you earlier to shape the compression into sounding smoother, or a bit more "natural". But know that these curves vary from compressor to compressor so 200ms on one compressor can sound exactly the same as 100ms on a different compressor.

Squashing It - AKA "Fat Drums"

We've seen how to make a "Fat" drum hit snappy again by delaying the Attack a bit before compressing the hit... But what about making a snappy hit into a big "Fat" hit? Well thats easy, what you want to do is use an Extremely fast Attack time (fastest possible) but dont use a fast Release along with it or you'll get distortion, set the Release atleast above 50ms-100ms, you can use long Release times so the Attack stage doesnt have to do all the work, the Release stage will then be able to compress alot of the peaks as well with a long Release time. Place the Threshold where you want the compressor to start chopping away at those thin peaks. Use a high Ratio to make sure not much gets through, and then the trick is to use the Make-Up Gain to make it loud. Remember "Make-Up Gain" is nothing more than volume you add after compression.

Here is an example: in this image right here, I used the first breakbeat from the "101s Breakbeat Collection" called "Aerosmith - Walk This Way". This break has large peaks but the actual "power" of the break is not much, a bit on the thin side. So I used a compressor with the fastest possible Attack time, to not let any peaks through, a Threshold about midway between the highest peaks and the lowest peaks (-12dB), a high Ratio (7:1), and a medium Release (300ms) so the Attack doesnt have to do all the work, the Release stage will compress a few of those peaks as well; I didn't set the Release too long here, but you can play with the Release time to your preference. The last thing I did was, I ramped up the Make-Up Gain to make up for all those lost peaks, thus increasing the average volume. It sounded fatter than the original, but I wanted it even fatter than that, so I put another compressor in the chain and put in values a little less than the first compressor, here's what I came out with:


Some compressors might not be able to catch those first few peaks, or you might decide to delay the Attack a bit to let some of the original audio through, in this case you would want to follow the compressor up with a Limiter. A Limiter is somewhat like a compressor with the Threshold set at 0dB, the highest Ratio possible, the fastest Attack time possible and they usually have an "auto-release" function. A Limiter ensures that absolutely no peaks will go over 0dB (or whatever you set it to) to avoid clipping the audio and distortion.

Lets talk about the "Flavors" or "Colors" of Compressors

So above you saw how I was able to bring the very high peaks down and then turn the volume up higher, bringing the average of the volume up higher than it was before. This could be one of the reasons that some people look at compression as making things "fat", because it doesnt just boost the peak volume.. it levels the audio so you can then boost the average volume afterward, the RMS. But the average volume isn't the end of it when it comes to "that sound" produced by a compressor; Each compressor comes in it's own unique "flavor" or "color". Some compressors are known to sort-of "dull" or "soften" the audio, this can be more noticeably heard on audio such as drum-hits, you'll notice that the crack of a snare hit can become more "dull" and "punchy" with certain compressors instead of that sharp, raw, crack! of a snare. Other compressors are known to distort the audio, again fairly easy to notice on drum hits such as a kick drum, some people take advantage of compressors with this style of "flavor" and may use it for saturation purposes as well as bringin the average volume up. There is one other "flavor" of compressors, "the plain vanilla," the "transparent" compressor, the one that colors the audio very little but still lets you control the volume of the audio, this type of compressor does not distort and it does not dull the sound, it is "transparent."

FAQ about compression for beginners

User avatar
The Messiah of the 21st Century
The Messiah of the 21st Century
Posts: 9913
Joined: Sun Oct 28, 2001 12:00 am
Location: Stevenage


Post by DJ ABO » Thu Jul 29, 2010 10:35 pm

Equalizing for dummies! By Christian Zechner

OK, so you have worked for hours, days, weeks on this GREAT track of yours but no matter what you do, it just sounds flat or «muddy», you can't get the kick-drums to stand out, the vocals are almost impossible to interpret and it doesn't sound as clean, transparent and punchy as you hoped it would. Here I've tried to provide an interesting and useful read about the sound frequency spectrum in general, equalizers, filters and how to utilize these in your mix.

1. Introduction – the sound frequency spectrum
2. Music and frequency ranges – Equalizing
3. What is an equalizer?
4. Filter types
5. General frequency ranges
6. Helpful tips

1. Introduction – the sound frequency spectrum
What is frequency? If a digital audio signal has a frequency of 50 Hertz (50Hz) this tells us that the signal in question is cycling from its starting point (0) to positive amplitude to negative amplitude and back to the starting point 50 times per second. The lower the frequency is, the slower the signal will oscillate. (In synthesizers, oscillators with very low frequency settings are used as control parameters to control other aspects of a sound. These oscillators are called LOW FREQUENCY OSCILLATORS or LFOs.)

The OPTIMAL human ear picks up sound in the frequency range of 16Hz to 24 000Hz (24 kHz), but the average infant/adolescent person can hear sounds between 20Hz to 20kHz and the range for adults are generally 50Hz to 16kHz. This will gradually get narrower as you grow older. Some of you have probably heard of the popular ring-tone "mosquito" which exploits this fact. It's a very high-frequent sound and cannot be heard by most adults, so it is a very popular ring-tone used in classrooms, f.ex.. The upper limit of human hearing is caused by the middle ear acting as a LOW-PASS FILTER. If ultrasound is fed directly to the skull bone, much higher frequencies can be heard (<200kHz).

The INFRASONIC spectrum consists of sounds that have a frequency too low to be picked up by the human ear. This is everything between 0.001Hz up to approximately 20Hz. These sounds can often be felt physically, but not heard. These are very energy-loaded sounds, and they can travel great distances (after the great Krakatoa eruption in the 19th century, infrasound travelled seven times around the Earth) and through/around objects with ease. Earthquake- and tornado-warning systems and nuclear bomb monitoring systems utilize this.

Examples of this in nature are: sounds emitted during and prior to a volcanic eruption or earthquake, avalanches, ocean waves, tornadoes and other winds. It is also the preferred communication method by elephants, giraffes, alligators, rhinos and whales. Animals are thought to pick up infrasound prior to and during natural disasters, this is thought to be the case in 2004 when the great tsunami hit the shores of countries around the Indian Sea. The animals fled from the shores long before the tsunami hit.

Infrasound is also sometimes utilized in soundtracks and music, as it can produce feelings of fear, sorrow and anxiety when (unconsciously) detected by humans.

The ULTRASONIC spectrum, however, consists of the sounds with a frequency too high to be heard by the human ear (i.e. >20KHz). We've all seen how ultrasound is utilized in medical situations when examining internal organs is part of determining a diagnose, or to see the fetus inside a pregnant womb. Ultrasound is also used in dog whistles (16-22kHz) as dogs have a higher upper limit of hearing, and in sonar/echolocators on boats. Natural occurrences of ultrasound can be found f.ex. in the natural sonar used by whales and dolphins and the echolocation method used by bats.

2. Music and frequency ranges
In music, we will most of the time utilize the sonic range of 20Hz to 20 000Hz (most often written as 20kHz, or sometimes only 20k). We want to filter out any frequencies we can't hear to make room for and exploit those we CAN hear.

An equalizer can be explained as a volume control. But unlike a normal volume control (like the knob on your stereo system), an equalizer can be assigned to raise or lower the volume on a specific frequency instead of changing the overall volume. There are two types of equalizers, GRAPHICAL and PARAMETRIC. Parametric equalizers have a graphic preview of the EQ slope which most of the time can be directly tweaked by clicking on it, and you can assign any frequency and any bandwidth (also known as the «Q factor») to each band. Graphic equalizers on the other hand, have already assigned frequencies and bandwidths on each band and are therefore generally less controllable.

A filter is, as the name suggests, a method used when filtering out various frequencies. They are not as advanced and versatile as an equalizer, but the principles are the same. When you are working with an equalizer you can sometimes set various filter settings on different bands, most common on the lowest (left-most) and highest (right-most) frequency bands. This depends on what equalizer you are using. We have seven different types of filters, all of which can be used when doing only slight equalizing or as an effect. Different filter plugins will sound different, so experiment and find one that suits your taste. The filters are the following:

LOW-PASS (LP, also known as HIGH-CUT)
A low-pass filter will let through the frequencies BELOW the set cutoff frequency. This is the most commonly used filter type in electronic music, mainly as an effect. Electronic low-pass filters are used in subwoofers and other speakers, to filter out the higher frequencies that can't be translated well by the respective speaker. Useful as a simple EQ on bass sounds or for filtering out noise from bad recordings.

HIGH-PASS (HP, also known as LOW-CUT)
This is the opposite of a low-pass filter. A high-pass filter will let through all the frequencies ABOVE the chosen cutoff frequency. This is useful as a simple EQ on hi-hats/cymbals and other percussive drum sounds as well as for reducing/removing the lower frequencies in other non-bass instruments.

A band-pass filter will let through the frequencies on and around the selected cutoff frequency. Nice for percussive sounds and supporting basslines.

BAND-REJECT (also referred to as a BAND-STOP or NOTCH filter. A notch filter has a generally high Q factor)
Being the opposite of a band-pass filter, this will filter out the frequencies on and around the set cutoff frequency.

A peak filter is very similar to a band-reject filter, but in opposition to a band-reject filter a peak filter allows you to either boost OR cut on and around the specified frequency, rather than just cut.

Shelving filters are useful when you want to boost/cut every frequency below (low shelf) or above (high shelf) a certain frequency.

We also have the COMB filter (which are essentially several peak filters) and the FORMANT filter (>2 bandpass/peak filters on already assigned frequencies. This is more used as an effect rather than EQ so I will not cover it here).

Play with your favourite equalizer plugin and try out these various EQ settings. The following frequency range descriptions are meant to be used only as guidelines and are not to be followed literally.

16Hz – 60Hz = SUB BASS
This is the super low-end that can be felt physically by your body on a good subwoofer/sub-bass system. Sounds with these frequencies are the most powerful ones, and they will take up a lot of room in the mix. Use this range to fatten up your kick drums or sub-bass patches. Too much volume in this range makes your mix sound «muddy.»

60Hz – 250Hz = BASS
This is where basslines and kick drums have their most important sounds. A common problem is that the bassline and kick cancel each other out due to PHASE problems (easily demonstrated when DJ-ing, if you play two tracks and have them beatmatched, it's important to cut one of the tracks' bass level or else the kick drums will cancel each other out and the overall bass level is lowered). A useful trick then is to try PHASE INVERSION on either the bassline or the kick drum, compressing the kick and bass together and/or avoiding to place a bass note on top of a kick drum. This range should also be lowered in most other sounds like guitars, synth lines and vocals so they don't interfere with the kick and bassline. Too much volume here makes the mix sound «boomy.»

200Hz - 400Hz
Too much volume here will cause vocals to sound muddy and unclear. Cut this to thin out drum parts like snares, hi-hats, percussions and cymbals, boost to make them sound warmer or more «woody.»

250Hz – 2kHz = LOW MID or MID-LO
Most instruments have their «darkest» parts here; guitars, piano, synthlines. Boosting around 500Hz – 1kHz can sound «horn-like» while boosting 1kHz – 2kHz can sound metallic.

400Hz - 800Hz
You can reduce some of these frequencies on the master mix to make your overall bass level sound tighter. Boost or cut here to fatten up or thin out the low end of guitars, synthlines and vocals.

800Hz – 1kHz
Here you can also fatten up vocals and make them sound warmer, in a different way than the previously mentioned method. Boosting around 1kHz helps add to the «knocking» sound of a kick drum.

1kHz – 3kHz
This is the edgy part of a sound, boost (gently!) here to define guitars, pianos, vocals and add clarity to basslines. Cut here to remove painful mid-frequencies in vocals. This frequency range is very hard on the ears, so be careful not adding too much volume here!

2kHz – 4kHz = HIGH MID or MID-HI
Vocals have a lot of sound in this area, the sounds «B», «M» and «V» lie here.

3kHz – 6kHz = PRESENCE
Plucky, fingered guitars and basslines can be more defined by boosting in this range. Cut in the lower part to remove the hard sound of vocals. Cut in the upper part to soften/round off sounds, and boost to add more clarity or presence to a sound. Boosting here helps defining most instruments and vocals.

6kHz – 10kHz = HIGH
Boost this area to add more air and transparency to a sound. Crispness and and sparkle can be added by boosting this range on guitars, strings and synth sounds. Snares and bassdrums also benefits from boosting this area. In vocals, cut some of these frequencies (a de-esser plugin does this easily) to remove the hissing sounds. The sounds «S» and «T» lies between 6kHz and 8kHz and too much volume there will make the vocals stressful on your ears.

10kHz – 16kHz = HIGH
This frequency range is where the crispness and brightness of sounds lie, and hi-hats and cymbals are the dominant drum parts. You can boost here to add even more air and transparency to sounds, and cut here to remove noise and hissing sounds which is unwanted in a bassline, for example. Pads and atmospheric sounds benefits from a boost in this range to make them sound brighter. Be careful not to boost too heavily, or else the mix will sound noisy.

A rule of thumb is to remove unwanted frequencies before raising the levels of those you want, but remember: the more frequencies you raise in a sound, the harder it is to place in the mix.

90% of the time it's better to cut rather than boost frequencies in a sound.

When listening for a bad, sharp frequency (maybe your newly recorded fat guitar riff, or the beautiful vocal hook is really hard on the ears when listening on loud volume) a good tip to find that horrible frequency is to put on a peak filter with a very narrow bandwidth, high gain settings, and then sweep the peak filter across the frequency spectrum until you find the right spot where it sounds like knives are being stabbed in your ears (be careful with the speaker volume when doing this!). Now invert your gain settings to a minimum and cut this frequency. You can also increase the bandwidth a bit if it still sounds sharp. Another way is to apply a narrow peak filter, and start with the gain all the way DOWN (instead of raising it) and sweep across the spectrum until your instrument sits nicely in the mix.

It is really important to listen to the channel you are working on in both solo mode and together with the rest of the track. If it sounds weird in solo mode, it doesn't necessarily sound weird when played together with all the other tracks, so use your ears!

Assigning each instrument to its own frequency range will help making the mix sound clearer overall (e.g. Basslines to the low end, guitars to mid-lo and/or mid-hi, vocals to mid-hi/high frequencies).

You cannot modify a frequency in recorded audio if the frequency isn't present in the sound itself, so if you f.ex. have a «muddy» recording to start off with, removing frequencies can help making it sound clearer.

Removal of subsonic rumble from your mix is essential for getting a nice, transparent sound image (especially when you are mastering for a vinyl record. When writing audio data to a vinyl record, too much information applied will write a wide ridge in the record and the needle will get unstable and skip.) The application of a highpass filter will fix this. Set it to cut everything below 20-50Hz.

In general, the EQ slope on a master mix should look much like a «smile». Raise the bass and treble levels and/or cut the mid range. This will ensure a good low end while maintaining a good, clear treble, and the mix won't be too hard on your ears.

And lastly: if it's not broken, don't fix it!

I hope this was a useful lesson, please don't hesitate to ask if you want more tutorials/infos on music production! I'll try to do some more stuff in the future if you've found this helpful

(DISCLAIMER: This text is based on my own way and my own methods of working with equalizers, and it is not necessarily the RIGHT way to do it. Factual errors could occur in this text. Trial and error is, by no doubt, the best way of learning. This text is subject to change without notice.)

Computer Music
Digital Recording Techniques
... and my own brain