Kush Audio guru Gregory Scott addresses the most common misconceptions about compression.
Compressors… 99% of us use them and most of us have a solid understanding of what they do, when and how to use them, and, critically, how to get what we need from them.
But I’m going to make the bold assertion that very few of us truly understand how they work and exactly what all those controls do on a technical level – and that includes some of the people who design and manufacture hardware and software compressors.
This lack of understanding isn’t helped by a number of myths about how compressors work and how the controls on the front panel are able to do what they do which are perpetuated by bloggers, forum posters – even books and magazines.
This month I’m going to look at a handful of the most common myths.
Myth #1: Attack is the time it takes for a compressor to begin compressing once a signal crosses over the threshold
This may be my favorite audio myth of all time, because I think it’s the most pervasive. People who know a lot of things; people who’ve written books on the subject of recording and processing audio, have perpetuated this myth by writing about it, repeating it and passing it along as established fact so that you can read it passed off as fact in countless blogs and forums.
The only problem is that it’s completely, utterly incorrect.
in the pursuit of truth, justice, and all that is good in the world, I give you the correct definition of attack
And so, in the pursuit of truth, justice, and all that is good in the world, I give you the correct definition of attack:
Attack is the length of time it takes a compressor to apply roughly two-thirds of the targeted amount of gain reduction.
I say ‘roughly two-thirds’ because there is no agreed-upon, industry-accepted standard for what this spec actually is. Yes, you read that right: no two compressor designers will agree on exactly how to define, and therefore measure, attack. My definition above is within the ballpark of most thinking, so I’m running with it.
To understand this definition of attack better, you need to get some basics of compression established first. Let’s say your compressor is set with a threshold of -10dB and a ratio of 3:1. If you feed this compressor a signal at -11dB, nothing happens because the signal is lower than the -10dB threshold.
But if that signal jumps to -1dB things get interesting. Most notably, the instant the signal reaches -10dB the compressor begins attacking it. There is no delay whatsoever in this response, which belies the myth that attack is the time it takes a compressor to respond once a signal crosses threshold.
With a -1dB signal and a -10dB threshold, the signal is 9dB over threshold. Our 3:1 ratio means that for every 3dB coming in over threshold, the comp wants to allow 1dB out the backside. Since our example has a signal 9dB over threshold, our hypothetical 3:1 comp wants to compress those incoming 9dB into 3dB at the output, which would require 6dB of gain reduction.
Given that attack is the time it takes a compressor to apply roughly 2/3 of the targeted gain reduction, the attack in this case indicates how fast the comp will apply the first 4dB of the target 6dB of reduction.
If you don’t follow the math of this illustration, don’t worry. For now it’s enough to know that the compressor starts applying gain reduction as soon as the signal crosses the threshold. Which means that attack is not a delay before action, nor is it even a measurement of time per se; instead, it is a rate, a measurement of the speed at which the process of gain reduction is occurring.
Myth #2: Release is the time it takes a compressor to release compression after the signal drops below threshold
Without going into detail, let me just say that the above definition is not only incorrect – it would actually be an impossible thing to assign a single value to. (Which is a story for another column.)
The correct definition of release will come as no surprise given what you’ve read above:
Release is the time it takes a compressor to restore two-thirds of the reduced gain to the compressed signal.
‘Restoring reduced gain’ is a very carefully chosen set of words. I characterised release in those terms because it’s useful to think of compression as a two-way street.
it’s useful to think of compression as a two-way street.
When a compressor attacks, it is applying gain reduction – it is lowering the signal level.
But gain reduction is only half the picture, because for every dB of gain a compressor takes away, at some point it has to put it back. And that process – let’s call it ‘gain restoration’ – is the business of release. The faster your release, the faster the compressor restores the gain it took away when attacking.
So what do we know now, at least in a purely academic way?
Attack is the length of time it takes a compressor to apply roughly two-thirds of the targeted gain reduction.
Release is the length of time it takes a compressor to restore roughly two-thirds of that reduced gain.
This gives us a good grounding to tackle more compression myths.
Myth #3: A compressor won’t release until the signal drops below the threshold
If you’ve been paying attention, it should already be obvious why this statement is false.
The explanation lies in the fact that aside from generating ancillary effects like distortion and colouration from transformers and tubes, attacking and releasing a signal are the only two things a compressor can do.
Put a little differently: any time the gain reduction meter on a compressor is moving, it is either attacking or releasing the signal.
Fascinating! Taking it a step further:
Any time the gain reduction meter is increasing (i.e., the comp is reducing the gain of the signal), the compressor is attacking.
Any time the gain reduction meter is decreasing (i.e., the comp is restoring the gain of the signal), the compressor is releasing.
So while the well-intentioned myth-spreaders out there would have you believe that attack and release are only relevant when a signal crosses the threshold – attack on the way up and release on the way down – what I am telling you is that nothing could be further from the truth. Instead, once a signal is over the threshold, both attack and release are constantly at play.
Any time the gain reduction meter is increasing, the compressor is attacking.
There’s a simple way to confirm this. Feed a drum loop into a compressor and set it up so that the signal is always over threshold and the gain reduction meter is dancing between say 6 and 12dB of reduction. In this instance the compressor is constantly attacking and releasing the signal, as indicated by the dance of the meter.
If the myths were true – if attack only happened when a signal crosses above threshold, and release only happened when a signal drops below threshold – adjusting the attack and release knobs in the above scenario wouldn’t make any difference because the signal is perpetually over the threshold … but turn the attack and release knobs and you will very clearly hear the sound of the continuous compression changing. Give it a try.
I think most people who use compressors on a regular basis already understand the above on an intuitive level, but some never make the connection that the behaviors they’re hearing (and seeing on the meters) don’t comport with the conventional – and flawed – wisdom.
05.08 PM
Really interesting article. I’ll need to come back and read Myth 5 again tomorrow to fully understand it, but that myth was the most interesting idea: “mixing is a game of balances, of relentless tradeoffs and compromises.
You don’t want every sound to be as dense as possible”.
Something to consider. Nice one!
05.22 PM
Amazing article! Thank you!
05.56 PM
Here I went thinking this was another dumb article on compression, after reading countless articles that basically re-write the manual for every compressor out there. This was great though. It really got me thinking about compression differently. A few ideas were only vaguely acknowledged in the back of my head until now, like the attack and release happening all the time once over the threshold.
Nice analogies and reference to density too. I’ll definitely be listening differently now.
02.24 AM
As soon as I see Gregory Scott’s name on an article I know I am in for a good read and always learn something. I love the studio tours here at attack but I am starting to really look forward to the articles by Mr. Scott …. not sure on my favorite section at this point. Mr. G Scott should really write a book on Studio Technologies.
02.42 AM
Exactly what I have always felt, but only subconsciously… Yeah, so many people don’t get compression, I hear it all the time. Myth 5 is so big. I was beginning to believe people who say “it sounds big aye?”, when it really just sounds more dense; squished of its (potentially big) life… Anyway, I couldn’t have put it into better words than Mr. Scott. Good article!
03.16 AM
Your explanation of ratio is backwards. For every 3dbdb of input over threshold you get 1db of output over threshold. The way you said it 3:1 would be more gentle than 2:1.
03.25 AM
Was literally just about to say what Pete said. It should be said “for everly 3dB the input signal passes the thresold, only 1dB will be allowed to pass by the compressor”. This article is great though!
03.37 AM
*threshold
03.43 AM
i loved this article but i was over here thinking i was losing my mind over what he said about ratios. glad someone else noticed that.
03.59 AM
Crap not to be a stooge, but also noticed this wording “if your attack is slow enough, the loudest bit of that transient will come screaming through before the detector tells the gain circuit ‘TURN IT DOWN!” I think saying that it will come through before the detector tells it to turn down, confuses people back to thinking that “Attack is the time it takes for a compressor to begin compressing once a signal crosses over the threshold” maybe worded like this “if your attack is slow enough, the loudest bit of that transient will come screaming through in the time that it takes the compressor to ‘TURN IT DOWN!” to the targeted gain amount” fits the point of myth #1 better.
Only pointing these out because i like this article so much and really wanna share it with everyone who i know struggles with years of people/the internet explaining how compression works in a slightly wrong way.
06.35 AM
Peter, Kellen and others on our Facebook page,
Thanks for raising this. Gregory has corrected the column above to what it should have said.
07.11 AM
Yeah, thanks to everyone for giving me my best facepalm moment of 2015!
Compression guru my ass, LOL!
08.40 AM
Love it this article! One of the best I’ve seen about comps. Greg, why don’t you do a video tutorial series about compression techniques?, that would be epic!
10.20 AM
This is a fantastic article! I teach students at university and those compressors always are most mysterious creatures for them. It’s that mix-up of time and attenuation levels that gets everyone (and got me too for ages).
A possible explanation for the attack/release confusion maybe that engineers learn the concept of attack and release in relation to sound shaping envelope in synthesizers, where those parameters are related to time… So automatically they transfer the idea to compressor when they see those settings on the box/screen. But it is not like that.
When using a compressor I have a vision of a particularly attentive and fit engineer with his hand on a sprung lever. The second he hears a signal louder then the set threshold he presses the lever down… The louder the sound – the harder he presses and holds it till the signal starts decreasing, then he starts releasing… Note, all of it is above the threshold. Time comes into it in some way, but it’s the pressure of the signal that is more important and how quick is the engineer… Cause if he is set a bit ‘slow on the uptake’ – those fast drums transients peaks will certainly get through unattended…
Oh and good point from Kellen above about drums transients – that one sentence got me a little confused too. Otherwise this is excellent. And the bit about the rubber band and the sound getting “bigger grrr” was an epiphany for me too. I will put it into our course reading and get on fixing my mixes!
11.53 AM
So refreshing to finally have an article to give my students as “post class reading / revision” that actually is the truth! I always feel a little cocky telling them that some books / online articles / etc… have it wrong, but this article will be my backup proof from now on!
12.04 PM
Lovely article – and I thought I was going to get to call-out the bit about “if your attack time is slow…” :-).
Comment about not applying gain reduction until the threshold is reached ignores knee settings – a soft kneed comp will apply GR before the threshold.
GR meters dancing above the threshold simply reflect that more or less level is hitting the key, so more or less level is being output – so you’ve got more or less gain reduction in terms of absolute level, but not in terms of reduction ratio (release knee notwithstanding).
Not picking – just a couple of points that could cause some confusion down the line.
02.24 PM
After reading the article, I believe Scott needs a hug. Someone please give him one. 🙂
02.26 PM
oops. “Mr. Scott”, as he is the teacher here.
04.35 AM
Interesting. I’ve lived long time with that springy and flexible thing but that size part is kinda new.
Also the attack part was new. I really thought there is delay. I was left wondering how soft and hard knee option in plugins affect that 2/3 timing.
Thank you. Great article.
06.58 PM
Great Article!
I would add one side note here that your Myth Busting descriptions are only fully true for Analog compressors because there was no escaping the limitations of the Physical components used.
Digital (plugin) compressors are not restricted by the same physical limitations and many use the Myths you describe as the Laws in their design. This includes most of the top manufactures and complicates defining things quite a bit. It also allows for many incredibly powerful tools that are impossible to create in the analog realm.
Michael Brauer’s description is right on point and defines the Art of compression as opposed to the use of compression.
Cheers!
Mike
12.08 AM
Where’s the proof??? Show us some math at least. Or is this just another completely subjective piece of tripe that ‘seems right’ because of the way it is written? Are there any standards left in the ‘pro’ audio world anymore?
12.16 AM
Sorr y- I wasn’t through when I hit submit. There is no account taken here of the knee, or the compression ratio and the interaction of the two at the onset of compression – at the very least. I hope the students who read this question it instead of taking it as ‘gospel’
10.54 AM
John Paterno: myths 1-4 are fact, only 5 is subjective. I’m unclear whether you actually read the whole article, it doesn’t seem so.
You ask (albeit rhetorically) whether there are any standards left in the pro audio world; ironically, there is no agreed upon standard for exactly how to measure (or even define) attack, and to the best of my knowledge there never has been.
It doesn’t matter what the ratio and/or knee are, attack is what it is. A steeper ratio and harder knee (e.g.) simply means that, with an attack of Xms, more compression will be applied within those Xms than would if the attack were slower and the knee softer. Same attack, same signal, different GR occurring in the same slice of time.
04.53 PM
I read it. Twice. There is no description of the methods you used to come up with these conclusions. There is no ‘academic’ method here at all. How are you testing these assumptions? What is your method? How are you measuring it? You are creating a definition of attack but not backing it up with any rigorous proof, thus creating your own myth in the process. – but trying to pass it off as fact. Kinda like Donald Trump. My issue is not that you have an opinion on how it works, but that you are stating things as ‘facts’ when you can’t even narrow in on a number. ‘Roughly 2/3rds’?? Is that a fact?…
05.29 PM
You ask “‘Roughly 2/3rds’?? Is that a fact?…”
I’m hoping that’s a rhetorical question, if not the word ‘roughly’ should give you the answer, but I get the feeling from your approach you are not looking for answers but an argument.
07.39 PM
I’m looking for the relevance in practical terms and application, and why the ‘smoking gun’ hype-riduculousness in the tone of the presentation. The RC time constant, and T, are phenomenon that designers have had to deal with since they coupled a resistor and a capacitor together.
For those interested in what he’s really talking about – which he failed to mention – from Wikipedia ‘RC Time Constants’ :
T is the time required to charge the capacitor, through the resistor, by ≈ 63.2 percent of the difference between the initial value and final value or discharge the capacitor to ≈36.8 percent. This value is derived from the mathematical constant e, specifically 1-e^{-1}, more specifically as voltage to charge the capacitor versus time
Charging V(t) = V_0(1-e^{-t/ \tau}) [1]
Discharging V(t) = V_0(e^{-t/ \tau})
So yes, this phenomenon exists, but to say the rest of the world is wrong in practical usage is pretty arrogant at best. He’s not ‘demolishing a myth’, he’s focusing on the minutia that even real designers do not worry about by the time their product hits the shelves. Is it something designers struggle with? yes. That’s precisely WHY there is no standard definition. Does it affect the practical application and end use. No. ‘Completely and utterly incorrect’? Seems to me it is more splitting hairs.
In full disclosure, it’s been a long time since my college electric engineering classes, so I consulted a couple of designer friends who pointed me in the right direction with this ‘roughly 2/3rds’ business. But I at least have tried to explain where that number comes from here, even if the author did not.
08.07 AM
“Seems to me it is more splitting hairs.”
Wow, just wow! Your lack of self awareness is astounding.
01.17 PM
Give it a rest , Ness. No need to be a jerk.
Besides, aren’t our ears the most important things when it comes to audio engineering? As valid as the maths surrounding compression are, we should always remember to challenge ourselves with questions like ” Does this sound good?” Which seems to be Greg’s point, hence myth 5. Of course that’s subjective so no point debating what “good” means.
Good article, Attack. I’d be stoked to read Greg’s opinions on EQ or EQ vs compression in mastering in future.
09.54 AM
Thnx mr Gregory Scott for the subject, that to Michael Brauer for the rubber “thing” and thank you John Patterno for the capacitors, also i would like to say thank you to Mike who has left the comment and explained a bit in his video about this subject when it comes to analog and digital compressors and why this can not be applied to every digital comp.
10.49 AM
Great article! I am not sure if I understood the following correctly :
“Any time the gain reduction meter is increasing (i.e., the comp is reducing the gain of the signal), the compressor is attacking.
Any time the gain reduction meter is decreasing (i.e., the comp is restoring the gain of the signal), the compressor is releasing.”
Do you mean that, for every db of gain reduction added over threshold, the attack time dictates how fast the compressor is adding it ? ie. a 20ms attack time would affect how fast a compressor goes from a -5db of gain reduction to a -10db one ?
I always thought that once the threshold was crossed, there was some sort of linearity in the amount of gain reduction applied (with the ratio affecting the GR with, well I guess, some sort of look-ahead technology).
06.13 AM
Great info…in the comments too.
02.05 PM
Very interesting, but I don’t think having a faulty understanding of exactly when the attack kicks in really changes how anyone will use a compressor. Slower attack times = more transients get through. That applies whether the attack time controls how fast the compressor starts working or how fast it achieves 2/3 of the desired gain reduction.
I understand from an electrical engineering standpoint the difference is important. But from a “let’s creative beautiful music” standpoint, it doesn’t really change how I use a compressor when I reach for it.
11.57 AM
@ Joe Gilder: indeed the creative process isn’t hindered by not knowing this (as you have proven numerous times). But in my opinion insight in the technical processes behind the compression will sometimes get you where you want to be faster.
It helped me dealing with my drum tracks and it felt “right” to me for understanding the technical side.
06.53 PM
brilliant article! would love to see one in a similar vain on gates/expanders. lately I’ve been using them a lot in my productions and they seem a lot more relevant than compressors oddly enough
04.13 PM
I still dont have any idea how to use a compressor, but I feel a lot smarter
07.05 AM
Nothing new, every compressor spec sheet will tell you this.
09.51 AM
Thank you Greg.
09.08 AM
I was always taught to set a fast attack for parallel compression. As a matter of fact there is an article by Bruce Aisher for Attack Magazine where his suggestion is, “attack time should be as fast as possible”.
Can we clear this up?
05.42 PM
awesome article, helped me a lot..thanks!
any chance of getting an article on WHEN to use compression? or anyone have any good links?
thanks again!
11.26 PM
@ John Paterno Maybe give it a rest. Greg’s definition of attack is spot on.
He doesn’t need to follow the “scientific method” (LOL) to give his explanation, as he is not investigating some phenomenon, just reporting on the facts — namely, the standard practice of compressor makers. It’s not like you need to measure it to verify it’s true, you can always ask any electrical engineer working in analog compressors or VST programmer modelling them.