Average and ultimate number of compression stages-What are your thoughts?

Answers will vary as much as estimates of stars in the sky. I, too, have wondered about how many compressors to use. I also wonder if my use of compressors is compensating for bad technique somewhere else.

I record my lead Vox on two channels - one DI and one through a tube preamp. The DI is compressed heavily with boosted treble. The Tube is treated with a general compression.

The compressor on the group sum of those two channels is another compressor that I use to "finalize" things. The other two compressors were tailored to those track's needs. This third one is the "real" compressor that makes the necessary changes so the vox will sit in the mix.

My acoustics have a compressor on each one of them.

My electrics have compression, either because of the distortion or I add one on clean tracks.

My bass is compressed.

My drums are NOT, but I use Addictive Drums, and those guys know better than me.

I use NO compression on the bus for the studio master.

In Wavelab, for mastering, I use the API-2500 Waves VST for a 1.5 ratio just to catch the peaks. The needles barely twiddle on that.

So, I use a lot of compressors. Hey pros, am I using too much?

I've used two on a track: first one catches peaks and second one for a more general compression effect.

Also I've used a send to send the signal to another compressor to put a heavy squeeze on it while keeping the original uncompressed. It's called parallel compression, also "New York Compression":

https://secure.wikimedia.org/wikipedia/en/wiki/Parallel_compression

And the exciting compressor, which is similar but squeezes the treble a lot:

http://www.recordinginstitute.com/R2KREQ/excomp.htm

I've been using a lot of vintage-style compressor plugins, so the eq curve of the plug is as important as the amount of compression given at each stage (which is almost never more than a couple of dB for me). In the CD I'm working on now I have the UAD Studer tape emulator followed by an additional compressor plugin on every track. Different compressors or each track - LA2A or the Fairchild on things that need fairly gentle compression, 1176 or LA3A if I want something more aggressive. Still only a few dB of gain reduction at any stage.

I don't compress my drum or vocal buses - at least I have not done so on the recent CD. I'll limit the master bus if I'm going to listen to it compared to reference material before being sent out to be mastered.

With vintage compressors there wasn't much choice of ratio or attack and release times. So you can't really set them close. To me aggressive means higher ratios, harder knee, and (relatively) faster attack and release.

BobRogers, post: 360530 wrote: I've been using a lot of vintage-style compressor plugins, so the eq curve of the plug is as important as the amount of compression given at each stage (which is almost never more than a couple of dB for me). In the CD I'm working on now I have the UAD Studer tape emulator followed by an additional compressor plugin on every track. Different compressors or each track - LA2A or the Fairchild on things that need fairly gentle compression, 1176 or LA3A if I want something more aggressive. Still only a few dB of gain reduction at any stage.

I don't compress my drum or vocal buses - at least I have not done so on the recent CD. I'll limit the master bus if I'm going to listen to it compared to reference material before being sent out to be mastered.

I'm about to buy one of their UAD2Laptop pieces, was looking at that Studer emulator as well yesterday, looks like a lot of fun. How're you liking it so far?

Jmm22 - I was about to say what BobRogers said, but then he said it first! A lot of those vintage comps (and plugins based on them) have almost no controls at all beyond more or less compression and some really basic level controls.

EDIT: Also, I routinely use multiple comps per channel on vox, drums and bass. I don't often use more than 2 or 3, and when I do I attribute it to my own inexperience and lack of skill, as I should be able to get better results with less layers of comp I believe.

jmm22, post: 360532 wrote:
My intuition is that if one is going to stack compression in any sense, the compressors should be disimilar. Is this a fair generalization?

Sorry for the double post, but I wanted to address this - the answer is NO. If you have vintage emulating plugins, then maybe yes, because one comp might be faster for knocking off those crazy peaks, and then one might be slower and can even out the rest of the track's volume, so in that case 2 different comps is a good idea, and often necessary. BUT - if you only have one type of compressor plugin you can still mimick this effect, you stack 2 of them and set the first one to faster attack/release, higher threshold (and maybe higher ratio, depends) to handle severe peaks, and then a second one with slower attack/release, probably lower threshold, and possibly lower ratio (huge generalizations there... take with salt).

So you don't necessarily need different plugins, just different settings if your plugin gives you enough control.

I find that compressing compression invites a whole 'nother set of problems to a single track.

+1

I thought it was fairly standard practice (not all the time, but often anyways) to use a fast comp in concert with a slower one, especially on percussion tracks?

Only if one of those is serving as a limiter.

Ok, so basically what I'd outlined earlier, fast attack and release for the first comp, higher threshold and (possibly) higher ratio? Just so I'm understanding right, you're saying that it only works well if that ratio is set to infinity (or we'll say whatever the max is for that comp), or you're saying that whole technique works regardless of ratio because one is just targeting the high peaks with the first comp?

Not trying to be obtuse, just trying to make sure I understand!

AToE, post: 360535 wrote: ...I'm about to buy one of their UAD2Laptop pieces, was looking at that Studer emulator as well yesterday, looks like a lot of fun. How're you liking it so far?...

I like it quite a lot. It's really making this folk/country project I've been working on come together. By choosing tape type, speed, and bias you choose a mild eq curve that gives a cohesive feel to the songs. Controlling the input to each track adds a bit of tape compression to those tracks that need it. I certainly don't slam it, so a "real" compressor in the signal chain helps.

The UAD2 Laptop solo comes with the LA2A and the 1176. (The vintage style compressors are what attracted me to the UAD-2 in the first place.) I use those a lot and you can get a lot of instances of them - even with the solo. The Studer is just out, so it is at its maximum price right now. Also, it eats a fair amount of DSP, so it may not be the most practical plugin for a solo. (Though you could put it on a track and bounce down and then turn the plug off.) You will get 14 day demos on each plug which you can activate at any time. Pay attention to how much DSP each plug uses and develop a strategy for using the solo.

Davedog, post: 360537 wrote: I find that compressing compression invites a whole 'nother set of problems to a single track.

I thought it was fairly standard practice (not all the time, but often anyways) to use a fast comp in concert with a slower one, especially on percussion tracks?

These are not contradictory statements.

The problems are dynamic. You can get some crazy nonconvex gain curves with two compressors. The practice of using a compressor with a slow attack and a low ratio at one threshold and one with a fast attack and a high ratio (e.g. a limiter) at a higher threshold is one that works to give a nice double knee gain curve that permits a controlled attack but a lot of punch. It's definitely possible to add a lot of problems, but if you keep it simple its not so bad.

jmm22, post: 360555 wrote: But would it be fair to say...

viz. The Beatles:

Rules:
1. DO whatever sounds good.

2. DON'T blow up the equipment.

3. OBSERVE these rules in the order written.

(Of course, they could afford to blow a few things up. I can't LOL!)

:D

What is the point of your compression in the first place if you are layering two or three instances? I think the whole point of compression is not understood if this is in fact the case. There are specific reasons (ie limiting) where two instances might happen on the two bus or on a snare. In general practice it would be better to ask yourself what you are trying to achieve and use it once. If the initial track is properly recorded in the first place, multiple instances of compression shouldn't be necessary. Remember, a double edged sword cuts both ways.

TheRealShotgun, post: 176108 wrote:
Shotgun's Compressor Tools 1 of 2
What you have to do is understand what compressors do, and what each of the controls do IN GENERAL. Then you apply that knowlege to what you want out of using the compressor and what your ears hear AT THE TIME OF USE so that you can adjust as necessary. So, read below for an overview of the box as a whole and each knob you're likely to find on it.

Compression
From the name, one can surmise that a compressor is going to squish, squash, mash or pulverize something. Given that we plug audio signals into it, we can further surmise that what is getting squished, squashed, mashed or pulverized is, indeed, our audio signal. And one would be completely correct in assuming that. But what does that really mean?

Well, consider an audio signal. Let's say it's a recording of my mom yelling at me about leaving my laundry piled haphazardly in the hallway. First, mom starts out trying to reason with me, gently, "Shotgun, you know, it's just not condusive to laundry efficiency leaving that stuff piled haphazardly like that..." her voice is calm, even and even somewhat soft. As I stare at her blankly, not understanding the finer points of sorting one's laundry and transporting it to the appropriate room in the house her voice becomes stronger and louder. "SHOTGUN! I'M GOING TO BEAT THE LIVING SHIT OUT OF YOU WITH A TIRE IRON IF YOU DON'T PICK THIS SHIT UP IMMEDIATELY AND PUT IT WHERE IT BELONGS SO HELP ME GOD!" Now she's yelling, screaming, in fact. Her face is red and frankly, I've just soiled myself which makes the entire laundry issue even more complicated.

Now, let's assume we're going to lay this recording of mom over some Nine Inch Nails-style door slamming, pipe clanging, fuzz guitar backing tracks. It's going to be an artistic tour-de-force. However, when mom starts out, her voice was hitting only about 65-70dB--normal conversational speech. By the time she's done it's more like 105dB worth of banshee howling. Unfortunately, our backing tracks are a pretty even volume the whole way through. So, at the beginning of the track mom will be virtually inaduible whereas at the end she'll be drowning out my samples of whacking a stapler on a desk. How do we deal with that?

WE USE A COMPRESSOR!

You see, what a compressor compresses is volume. That is, technically, it compresses the amplitude of the signal, or its "gain". So for every decibel that goes into the compressor, only a fraction of it will come out. That means that (depending on our settings, see below) if mom's voice uncompressed winds up at 105dB then we can set our compressor so that it only gets as high as 52dB if we want. How does that help you ask? Won't it still be too low to hear over the backing music? Yes it will, but read on and we'll cover that in the controls discussion.

Threshold
The threshold control on a compressor sets a level below which the compressor will do no work. The control is graduated in dB (in this case dBV of signal level) and allows you to set an "on/off" point so that you can compress the LOUD parts of a signal, and leave the soft parts alone. At times you may want to set this control low enough so that you're affecting the entire signal, at times you may not. In the case of mom's rant-on-tape, what we may want to do is set the compressor so that it doesn't touch the signal until her voice reaches something like 85dB or so***, say, about halfway up the scale from softest to loudest. So, we set the threshold so that we only see activity on our "gain reduction" meter when the track gets to a certain point.

To USE the threshold control effectively, you generally need to use your ears. Have some idea, before you start, of what you hope to accomplish by using the compressor and set the threshold to capture the part of a signal you wish to do whatever that is to. In our example I want to lower the louder parts of my mom's tirade so I set the threshold to activate the compression at some arbitrary point in the track. I could have done it several other ways and the only way to learn which is best is to experiment and listen.

Ratio
This is the control that tells us how much signal comes out of the box relative to what's coming in. It is graduated in terms of a ratio (hence the name) of output to input. So, let's say we set the control to point at "2:1". That means that for every 2dB of incoming signal, we're only going to get 1dB of outgoing signal. Which means that at its very loudest, mom's voice isn't going to be nearly as loud as it was originally. Keep in mind that this ratio only applies to signals that meet or exceed the threshold setting. Any signal that is below the threshold just passes through as though the compressor weren't there (kinda).

To use the ratio control effectively you, again, need some idea of what you want out of your compressor overall. In our case I just need mom's voice to be more easily mixed in with the backing tracks so I just want it to be kind of even. However, I still want it to start softer and get louder, just maybe not AS soft at the beginning and not AS loud at the end. That is, still changing, just not as much.

Attack
The attack control tells us that, once a signal meets or exceeds the threshold, how quickly does the compressor put the smack down on said signal? The control is usually graduated in intervals of time, usually marked in milliseconds. So, let's say that I set my attack control to say 5ms. That means that when the signal passing through reaches the threshold I've set, the compressor waits an additional 5ms before it begins to reduce the amplitude (again, gain). This seems counter-intuitive doesn't it? I mean, we want the level controlled WHEN it reachest threshold, right? Not 5ms later. Well, there are reasons for slightly delaying the attack (and for that matter release) times.

To use the attack time effectively (and by now you should have seen this coming) you need to know what you want out of your compressor in general. Do you want the signal clamped down on fairly quickly? Or not? How do you know? This brings in one of the most important concepts of recording: attack and decay. Each sound has an attack and a release. Imagine hitting a drum (the easiest place to see this concept). You hear the sharp, immediately loud sound as the stick hits the head, but you also hear the sound gently fade away, also. That initial WHACK, that initial spike in amplitude is the sound's attack. Everything else is it's decay. Note that I use these terms in a "Shotgun" type of way and there are more correct ways to say this, I think, but I tend to, over time, develop my own language, so you're at a disadvantage.

So then, we can hear an attack in mom's voice, too. It's more subtle than the attack of a drum hit with a stick, or a guitar player's pick against a string, but it's there. And if we set our compressor's attack time too short, we will lose all the definition of the attack of the sound. Sometimes that's desirable, but in our case it is not. A very large percentage of how people perceive sounds comes from the attack. You must strive to preserve that unless it is your desire to purposely not. Therefore, be very careful with the attacks under your care. In the case of a vocal track, the attack of the voice will lend very much to the intelligibility of the track, so we do NOT want to destroy it. So, we may want a slightly longer attack time than 5ms here. But we can only tell BY LISTENING. LISTEN to the track, sweep the attack control back and forth and listen to what happens to the attack of the sounds. If it sucks, move the control. Don't look at where it's pointing until you're satisfied with how it sounds. Then only look for the sake of curiosity because that setting may never work the same way again. if you're using a plug-in make sure you allow ample time for the movement to take effect. Moving a plug-in's controls can sometimes not take effect for a full second or two after you move it so if you're sweeping it back and forth rapidly you'll fool yourself. In the case of plugins, make a move and pause until it changes. If it doesn't change within 2-3 seconds, maybe you didn't move it far enough.

Release
As you might guess the release control handles the other end of the signal from the attack. That is, when a signal drops back below the threshold, how long does the compressor wait to actually stop compressing. All the same counter-intuitiveness applies here as well. However, remember that the decay or "tail" of a signal isn't as important to the listener as the attack so you can get away with a little more here. Again this control is going to be graduated in units of time, usually ms. However, the numbers will be larger than the attack times. Sometimes up into the 100's of ms or even full seconds.

To set a proper release time, again, understand what you want out of your compressor. Do you want a major thrashing to your sound, or do you just want kind of a gentle corrective measure? What you have to look out for in the case of release times is pumping. If your release time is set too short then the sound will drop below the threshold, the compressor will release it, but the sound will then jump UP in level because the compression is no longer making it softer, but it's below threshold. That probably sounds confusing, but it happens. And it will sound pretty odd. The first time you hear it you'll understand why it's called "pumping". It sounds almost like there's a new "attack" near the end of the signal's decay. As I've said before, sometimes this is actually desirable. Usually it's not though. Your goal is to set a release time long enough to give the sound time to naturally decay to a point that when the compressor lets go it won't "pump" yet short enough so that the compressor isn't still active when the next "attack" comes along. If you set your release time too long it will start ****ing around with the attacks because it's taking so long to let go the next loud signal is there before the last one is finished compressing. So, if you get your attack set where you think it's right, but then you start losing your attack again, consider dropping that release time lower (faster).

Make up gain
Here's where we answer your initial question of "Won't it still be too low to hear over the backing music?" Remember that we noted that mom's voice started out so low that it was lost in the music. And all we've done so far is to use our compressor to take the bite out of the louder part of the track so that it's not overpowering. So, doesn't this leave the softer part still lost? And, possibly, doesn't it make the WHOLE TRACK too soft now? Yes, it absolutely does. But that's what we have makeup gain for.

The makeup gain is going to look very similar to any other gain control you have seen. It will be marked off in dB, possibly starting at 0dB and moving up to some obscene amount like 20 or 40 or 60 or 100,000 or something. (It won't really be 100,000). The makeup gain does just what it says it does, too. It allows you to "make up" the gain that you're losing by compressing in the first place. Now, that doesn't mean it UNDOES what you just did, not by any means. It means that you can now take your newly compressed signal and make the WHOLE THING louder. This is how we're going to get the parts that are too soft up where they belong.

To set this control we're going to, of course, listen. What we've done thus far is to compress down the loudest parts of the signal so that they're not so loud. You can say that the loud parts are now "closer" to the soft parts so to speak. So what you do with your makeup gain is to take the whole lot and move it back UP some smaller amount so that now the loudest parts are just still loud, but not AS loud and the softer parts are still soft, but loud enough to be heard. Think of yourself playing basketball. If you're short like me, there's no way you can slam dunk a basketball. However, let's say you can lower your basketball goal by one foot. Now it's lower, but you still can't slam dunk it, but lowering it any more would ruin the rest of the game because you'd just be dropping the thing in and not shooting. So what you do is you make yourself magically grow a foot as well. Now the goal is still a reasonable height, but you can slam dunk because you've grown a bit yourself. Same sorta thing. Your signal isn't so low it sucks now, but it isn't so high you can't get anything useful out of it as well.

Here's a shocker: in terms of makeup gain there IS a general rule you can keep in mind. As you're setting your compressor's other settings you will notice the meter marked "gain reduction" giving you some idea of what you're doing to the signal. It could be a schitzophrenic little peak meter or it could be a big, slow, thoughtful VU meter. Either way it'll tell you "hey, you're getting about 5dB of gain reduction here pal!" So, this tells you you can START your makeup gain at a setting of +5dB. That should give you a compressed signal at the same general level as the uncompressed signal. Kinda. Sorta. It's a really ROUGH starting point, but it's a starting point nonetheless. Again, though, twist it and listen to get it where it really needs to be. You may want more, you may want less.

The BS you'll hear
Now, as you get replies to this thread there will be plenty of numbskulls along to give the following answers:

(1) Shotgun you're such a ****ing asshole. The guy just wanted some basic info, some basic starting points for his compressor why do you have to be such a prick?

(2) Shotgun, you don't understand compression and you've never done any recording, HAVE you?

(3) Here are my basic settings and they'll probably work

None of that is even remotely true. Sure, there are plenty of basic starting points anybody here could give you. In fact, many of these folks have only been using compressors for about 6 months, but even THEY will have ONE setting group that they like for some reason and are DYING to tell you it in order to appear knowlegeable. Do not listen to any of this shit. Develop your own views on good starter compression settings by appying what you learn and what you hear and what you observe in your own experience. There are so many different kinds of compressors that anybody who gives you a rough setting diatribe is just pissing in the wind. In fact, many types of compressors don't even HAVE some of the controls I mentioned. Some have more. Also, there are plenty of points we haven't covered. For example limiting, which is a special kind of compression that uses a very high ratio (often infinity:1).

Software compressors are so expensive because it's a royal pain in the pussy to write that particular software. The stuff that a compressor does to a signal is ungodly complex. Not necessarily the "reduce gain by x when level reaches y, wait b ms and hold for r s.". That part is easy. The hard part is trying to model, in software, the "sound" of all the individual components of a compressor and how they interact with each other to product a "sound". You may be familiar with people talking about the "tube sound." Well, each electronic component is going to impart its own "flavor" onto the signal passing through it. Some compressors, in fact, are used not so much for compressing as they are for the mere sound imparted when signal passes through their components. For example, a studio I used to work at very often tracked vocals through an LA3 tube compressor set to very very low compression, hardly any audible compressing at all, just so that the vocal could pass through the tubes on the way in.

As to which of the two you use, well, it's up to you. One cannot say one way or the other is better without knowing many factors. Budget is one of them. There are several GREAT freeware compressor plugins available. Additionally, you may not have the facility to use hardware compressors effectively. It would require a mixing console, plenty of A/D and D/A and all the cabling and so forth to run a signal out of your DAW and into a compressor and back into your DAW (that is, I am assuming you're using a DAW since you ask about software compressors). On the other hand, many of the expensive, much sought-after hardware compressors have NO software equivalent and their "sound" can only be had by using hardware. (regardless of what that twit from Bombfactory says). The bottom line is that you should use whatever you can get your hands on because there's no such thing as too much experience.

Again, a judgement call. I assume you mean during or after TRACKING. Because, if you use compression after you press a few CDs of a mix, it's not going to do you much good. As I mentioned earlier plenty of people use some compression during tracking. Sometimes that's for the actual compression, sometimes just for the sound of the compressor's components but hardly any actual compression. The cardinal rule to remember here is that you can ALWAYS add compression if you feel like you need it and didn't use it tracking. However, it's next to impossible to REMOVE compression if you used it tracking and later decide you don't like it. When in doubt, track as raw as you can and add stuff later. That's way easier than trying to make up for a bad compression choice in the beginning. Furthermore, as a beginner, you're probably not going to have a good enough handle on what types of compression are going to work for the mix as a whole during tracking to really make a good judgement. It would be a shot in the dark at best.

~S

***IMPORTANT NOTE: Please understand that the dBSPL of mom's voice is NOT the same as the dBV of the signal to which we are applying compression. If we knew what type of microphone was used, and what type of recording medium we could make rough calculations as to what dBV we would be working at, but we don't, so just suffice it to say that mom's voice was soft, then very loud and that the signal starts off weak and becomes accordingly stronger. Just don't make the mistake of assigning any particular significance to the numbers I'm using, I picked them at random.

Obviously this is not my speech but is from the sticky at the top of this forum. The information is key and pertinent though IMO.

This is turning into a great thread, I think it actually has covered a few things Shotgun didn't talk about (oh yes, I've read it a few times, it's almost Slipperman-esque!).

I can see why 3 comps would start to be pointless, seems like there's little you could do with 3 that you couldn't do with 2 (as discussed previously, one of them effectively being a limiter) - but I know that I personally simply do not have the skill to deal with many signals with a single compressor. Especially kick and snare drums, which I find seperate tailoring of the attack and sustain to be very important to me getting the sound I want(though I fully admit that someone with more skill could quite possibly accomplish what I can and more with a single comp). Maybe I could get similar results from 2 different mics compressed in different ways, one dealing with attack one with the "tone" of the drum, but even then, if the drummer is less than perfect I can think of cases where 2 comps would save my butt!

Well.........if you think of a sine wave as if it were a woman's breast and the nipple is the part you are lopping off with the compressor it might make a better visual. Then lets say you made that nipple inverted by quite a bit because of the ratio of your compression. Now you hit it with another compressor with a different ratio and release curve. Now we have something visually that could have been painted by Salvadore Dali or Edvard Munch. It may be artful, but the intent better not be to be pleasantly appealing and rather something else.

I follow what TJA is saying, but maybe replace "sine wave" with "envelope" since yes, even on just one track the compressor will be seeing a lot more than a simple sine wave.

Oi. Try to use some eloquent imagery around here.  

Don't yous goys know how hahd it is fer us Lederneks to yuse r wrdz?!?

:wink:

Attached files

I miss the Shotgun.

When you use compression you are not only effecting the 'envelope' but also the sources dynamics. Thats what the release and make-up gains are all about.

An example of these effects can be demonstrated very easily.

Take an articulate and busy guitar lead as an example. Note how you can hear all the notes without any compression present. Even the 'Jazz' notes. Now compress the bejeeezuss out of it. Notice how the notes arent as articulate as before. Sorta like a Jimmy Page lead.....Now, without adding any makeup gain other than to bring it back to its relative volume in solo, drop it back into the music backing tracks........Where did it go?

So. Whats going to happen to this track when its limited across the 2-bus now? More disappearing? You bet.

Someone said earlier that if the tracks are recorded properly, compression just becomes an effect and not a repair. This is something to strive for.

Mr. Dutton! Drop you socks and grab your pencil! Stop thinking about breasts and nipples!

Gain curve: A wire has a gain curve of 1:1. One dB gain in the input yields one dB gain of output. Input-outpt graph is a straight line with slope 1. Suppose we put a compressor in the circuit and set the threshold to -20dB, ratio to 2:1, attack an d release time to 0ms (impossible, but bear with me). Input-output graph has a slope of 1 up to -20dB, a "knee" at that point and a slope of 1/2 (2 dB input to 1 dB output) after that. If we add makeup gain that just moves the whole curve up vertically.

What does that do to a drum hit. A drum hit has a sharp spike at the beginning and longer, smoother decay afterward. Well, everything above -20dB gets squashed to half it height. Good. That makes the decay flatter giving more boom after we add makeup gain. BUT it does the most damage to the spike at the beginning. Less spike - less high frequency crack. If we increase the attack time more of the spike sneaks through with more high frequency snap. Good. Ahh, but what if the spike now clips our signal after the makeup gain. Well we can add another compressor with a threshold of, say -6bd, a high ratio and a fact attack. Note that this puts another "knee" in our gain curve. The first at -20 dB and the second at -6db minus the makeup gain of the first compressor. If the ratio is infinite and the attack is 0ms then it simply clips the spike off at -6. A lower ratio makes for a rounded spike instead of a clip.

I'm sorry I ever used the term "nonconvex". It would make sense to mathematicians - and not all of them. It just makes things more confusing to everyone else. The easiest example of a compression artifact is "pumping" - background sounds that get noticeably louder when the main sounds pause. This can happen with one compressor set improperly. Worse things can happen with two.

BobRogers, post: 360581 wrote: Mr. Dutton! Drop you socks and grab your pencil! Stop thinking about breasts and nipples!

Gain curve: A wire has a gain curve of 1:1. One dB gain in the input yields one dB gain of output. Input-outpt graph is a straight line with slope 1. Suppose we put a compressor in the circuit and set the threshold to -20dB, ratio to 2:1, attack an d release time to 0ms (impossible, but bear with me). Input-output graph has a slope of 1 up to -20dB, a "knee" at that point and a slope of 1/2 (2 dB input to 1 dB output) after that. If we add makeup gain that just moves the whole curve up vertically.

What does that do to a drum hit. A drum hit has a sharp spike at the beginning and longer, smoother decay afterward. Well, everything above -20dB gets squashed to half it height. Good. That makes the decay flatter giving more boom after we add makeup gain. BUT it does the most damage to the spike at the beginning. Less spike - less high frequency crack. If we increase the attack time more of the spike sneaks through with more high frequency snap. Good. Ahh, but what if the spike now clips our signal after the makeup gain. Well we can add another compressor with a threshold of, say -6bd, a high ratio and a fact attack. Note that this puts another "knee" in our gain curve. The first at -20 dB and the second at -6db minus the makeup gain of the first compressor. If the ratio is infinite and the attack is 0ms then it simply clips the spike off at -6. A lower ratio makes for a rounded spike instead of a clip.

I usually have the comp acting as a peak limiter first in the chain, but I've noticed a couple people in this thread mentioning it in the chain after the "main" compressor... what would I be missing/adding by having it first rather than second?

EDIT: Also, non-convex made perfect sense to me if I pictured a graph with the horizontal axis as time and the vertical as amplitude. Maybe that was wrong though!

@Bob: You want me to stop thinking about whaaaa? Unner-edumakated Sergeant Dutton?

Mathematics r a gut subajeckt fur me. I unnerstant bullit trajectory 'n unnerlyin princi princi principalz uv elervation, wind und temper(ature) inter-acktion. Iz thar more toit?

TheJackAttack, post: 360563 wrote: ask yourself what you are trying to achieve.

This is the best advice IMO. Not just for compression, but for everything you do while mixing.

eg: lets say you are compressing a kick drum. You want to add some extra punch to the sound, but you also want to tame a few over-enthusiastic hits in places.

The chances are that your punchy (slow attack) compressor setting will make the occasional too-loud hits much worse. But setting a faster attack time to tame those peaks will kill the punch. Now its obvious that you need two compression stages: one with a higher threshold and fast attack to catch the occasional peak, and another with a lower threshold and a slower attack to add an overall punchiness to the whole part.