What is the deal with compressors?

YanKleber

YanKleber

Retired
This thread will certainly sound weird for you. Even the simplest things for you are a big mystery for me.

OK, my problem here is with compressors. I cannot to learn how to use them because I simply don't understand them in a conceptual manner.

I am sure that I have misunderstood compressors my whole life. All those years I have had a compressor as a tool to make a track sound leveled, by reducing loud parts and raising the quiet ones. However every time I try to use it for leveling a vocal track it seems not to do anything. I can see the led (or the needle) of the VU bouncing as if it was doing something but for my ears it doesn't seem to be doing anything. For leveling my vocals I have used volume automation and it surely does a job that I can notice. So the question that remains for me is... what the hell the compressor REALLY does and why I don't manage to use it?

:confused:
 
First, 90% of your volume control is going to be with manual leveling (or automated gross volume control). So you're not doing anything wrong.

As far as the compressor itself, with reasonably extreme settings, you should definitely hear level changes -- That said, think of it more as something that controls the dynamic range of the *signal* as opposed to just a volume controller.

Sure, it makes loud things quieter. But it also affects the swing of the signal itself.
 
At more extreme settings, you should be hearing a definite change in the dynamics of the signal. Make sure that you're using it as an insert effect and not on an aux send bus. You want 100% of your signal passing through the compressor, unless you're doing something more advanced like parallel compression. But for run-of-the-mill situations, add a compressor VST as an insert (in Reaper, that just means add it to the FX of each individual/group/folder track.

If you're not hearing the effect of a compressor VST, that makes me think that you're not hearing it at 100% wet as an insert effect, or that you might be using it as a send/return type of effect. That blends the original signal in with the compressed signal, and essentially defeats the purpose. The original dynamics would still be intact.

Or maybe you're not dialing it in "extreme" enough. You certainly don't want it to be obnoxious and devoid of all dynamic content. But maybe just for illustration purposes, turn down the threshold, turn up the ratio, shorten the attack time, and see if that starts making it sound squashed and compressed. Increase the release time and see if it starts sounding "pumpy". If not, then we may have some other issues to sort out.
 
You're not alone. I can't hear it working, either, OP. I heard that's normal and takes years to be able to hear it. I hope it happens at some point. For now as I wait I just put it on tracks that are known to be dynamic (like bass or voice) and dial it in so it's like -3db of gain reduction. I have very sensitive ears in general but can't hear compression for some reason. What sucks about not being able to hear it is I/you can't get creative with it.

I do hear it on the extreme settings (and it sounds like crap usually), but can't hear it if set to a normal threshold or ratio.
 
If you can find a video demonstrating the Avid/digidesign dyn3 compressor it might help.
I'm sure there are others like it, but the visual display on that comp makes it really easy to understand what's happening.

Of course you should use your ears and not your eyes but, for learning the basics, that might be useful.


Subtle compression is funny. I guess the idea is that you shouldn't hear it but you should hear the problem if you disable it.
Plenty of people use heavy handed compression sometimes for whatever reasons, though. That can be pretty obvious if you know what you're listening for.

Edit: I was really hoping there'd be a Seinfeld style punchline. :(
 
Compression is your friend, but (and a BIG BUT!) Only when used correctly.

By all means volume automate but compression is a little different, compression grabs the peaks that are so quick that volume automation is too difficult.
Sometimes you want to hear the compression, it is an effect as well as a volume leveller. That in your face snare sound would have some compression attached to it. The guitar sound that has a massive attack on the strum but also has a full body of volume after it, etc etc, all these things are dependent on compression.

I can tell you this, the settings can be explained over and over, but it's your ears that lets you know what is right and wrong when dealing with compression.

Laymen's terms:

Attack: how fast the compressor reacts to the volume rise, how long it takes for the compressor to start working
Release: how quickly the compressor releases after the signal volume reduces, how long it holds the onto the reduction
Ratio: the amount the compressor reduces the rate the incoming signal increases in volume (not turns the volume down) the volume will still increases but at a reduced rate, however if the ratio is set to infinity the compressor is now a limiter and the volume will not rise about the threshold (where the compressor starts to work).

Go back and revisit compression with these simple things in mind and experiment by putting a simple track through it (on its own) and see how it behaves with different settings.

Oh and there are no rules, I am always learning different approaches, often by accident.

Hope this may help

Alan.
 
When I taught music technology in college, I used to hate it when we got to compression. Some students could not detect it at all, even when set to silly levels. It's often described by imagining a very fast finger on a fader, levelling out the peaks and troughs, but that really only works with certain material. If you have a track with just kick and snare on it, as in thump crack, thump crack, the fader finger analogy works, twiddling the knobs on the compressor also has an audible effect. On a voice, or instrument, it's far more complex. In many cases I failed miserably to get them to hear it, especially when using mild compression, which is more common. Sticking a compressor on a mix really doesn't do anything radical enough to hear. The only thing that worked as a demo was using a bass guitar in a mix where it was finger played, not with a pick. The lack of definition often loses the thing in a busy mix. Adding compression, and making up a little gain lets it be heard without red lights coming on! I always used a quiet track with a solo flute, recorded quite low to let them hear pumping, as the low record level would bring up quite a lot of noise when the compressor worked, and you could hear the swoosh you noise as a note tailed off.

I'm not convinced that anyone ever heard the good aspects of compression, but they all heard them used badly, so they could avoid that at least. However, one student said "ok, I get it. We know how to hear a compressor used badly, and we know how to adjust them to stop them sounding bad, but if we can't hear a benefit, why do we have to use them?" I wonder if the answer too often is simply because we do? Some of the old vintage recordings that still get played lots didn't have compression used in the way we do today, and still sound good. However, the Phil Collins drum sound is really just a great effect, not meant to be natural.

In live sound for years the PA companies had racks of compressors patched across almost everything. Getting one of these to do a 60s tribute makes me laugh. Sound rule number 7 says do the sound check, get everyone happy, then switch on compressors and take fingers off faders. Hmmmm
 
If you can't hear the effects, maybe you're listening to the wrong parts of the track...listen for the loudest parts. If your reduction meter is kicking during the loudest sections, you should hear the volume decrease. To surely notice an effect and see what it's doing in more extreme circumstances, do this - lower the threshold to the floor (usually -48db) and then listen. If it doesn't sound ugly and lifeless, almost inaudibly probably, then you have something set improperly somewhere, perhaps your gain levels. try that first.
 
The reason to use compression, even if listeners think they don't hear it, is that human hearing has its own compression built in. There's a muscle called the tensor tympani that pulls on the tympanic membrane (eardrum) to protect it when it's subjected to higher sound pressure levels. The human ear parts evolved from jaw bones and muscles. If you hear a rumbling sound when you yawn, that's your tensor tympani muscles still reacting to the same nerve impulses as your jaw.

Compression acts as a cue, one of several means by which the brain can tell how loud something is. If you want to make a mix sound loud then compression can help to a point. Of course there are other things human brains use to determine loudness so what can be done with compression is, um, limited.
 
Well, ultimately the reason to use compression is to control the dynamic content of the source material for whatever reason. Sometimes that's to make it sound "loud", sometimes it's to fit it in the window between tape hiss (road noise, other instruments) and distortion. I think it really is not so important to hear it working properly as to be able to recognize the negative side effects of inappropriate settings. Course one man's negative side effect is another man's deliberate special effect...

The description of attack and release above are pretty standard, but I think a lot of people have a fundamental misunderstanding of those parameters. They are very much low-pass filters on the (usually rectified) detector circuit, but what they do is tell the thing how much (and how long) sample values that have come before impact the detected level now. So, it's not like a triggered ASR envelope where when the signal goes over the threshold it ramps up to maximum compression and then after it goes back below it ramps back down to none. It "looks" that way when you apply it to a steady-state signal, but it's not quite right.

Probably going too far with this, but the attack parameter slows down the action of the compressor at all times that the signal is above the threshold, likewise release should affect the detected level at all times when signal is below threshold. That's important, because the signal (well, anything but a pure square/rectangle wave) crosses the threshold anywhere from tens to ten thousands of times per second, and if the gain reduction actually followed all of that, it would be distortion.

Many digital compressors have an RMS time or window parameter which also tells the detector how much/many previous samples matter now. This is basically a low-pass filter that happens before the attack/release filters, and at a certain point can become redundant to them. But of course, this is still always looking in the past. The gain reduction in any compressor is never really giving you as much as this sample really should have. It's always playing catchup...

Unless it looks into the future. It's pretty easy in digital (as long as you're not trying to monitor live realtime input), but a bit more difficult in hardware. Some digital compressors call it lookahead, and some call it pre-comp or whatever. They all do the same thing which is to feed the detector circuit (and Apply RMS time, attack and release, etc) with the audio signal that is going to happen and apply the gain reduction that it calculates for that to the actual now-time sample that you're hearing.

Nowadays a lot of my compression has attack and release set to zero. I use pre-comp and RMS window instead. Especially for something like leveling a vocal - give it a really long pre-comp, set RMS window to about double that, doesnt take much actual ratio, but set the threshold so it's always doing some little bit. It's so transparently effective that it is very much like cheating.
 
ashcat_lt said:
The description of attack and release above are pretty standard, but I think a lot of people have a fundamental misunderstanding of those parameters. They are very much low-pass filters on the (usually rectified) detector circuit, but what they do is tell the thing how much (and how long) sample values that have come before impact the detected level now. So, it's not like a triggered ASR envelope where when the signal goes over the threshold it ramps up to maximum compression and then after it goes back below it ramps back down to none. It "looks" that way when you apply it to a steady-state signal, but it's not quite right.
Click to expand...

In the time it took me to research and formulate my disagreeing response I realized I didn't disagree.

ashcat_lt said:
Probably going too far with this, but the attack parameter slows down the action of the compressor at all times that the signal is above the threshold, likewise release should affect the detected level at all times when signal is below threshold.
Click to expand...

It's my understanding that the attack parameter controls the speed of gain reduction whenever the gain applied is greater than the target gain value produced by the detector/threshold/ratio components, and that the release does the same when the gain applied is less than the target value. Most of the release action happens above the threshold. The only time levels below the threshold are affected is when the signal falls through the threshold and the gain is still ramping back to 0.

ashcat_lt said:
That's important, because the signal (well, anything but a pure square/rectangle wave) crosses the threshold anywhere from tens to ten thousands of times per second, and if the gain reduction actually followed all of that, it would be distortion.
Click to expand...

Wouldn't it be more correct to say "above 10Hz" instead of "but a pure square/sine wave"?

This reminds me that when I was mixing live a lot I liked to use super fast release to make the compression more transparent. Someone on PSW pointed out exactly what you said, if the gain changes happen too fast what you have is distortion.
 
bouldersoundguy said:
It's my understanding that the attack parameter controls the speed of gain reduction whenever the gain applied is greater than the target gain value produced by the detector/threshold/ratio components, and that the release does the same when the gain applied is less than the target value. Most of the release action happens above the threshold. The only time levels below the threshold are affected is when the signal falls through the threshold and the gain is still ramping back to 0.
Click to expand...
I realized after I posted that the release actually is usually a ramp that has nothing to do with the input. If the input is below the threshold, the LPF is being fed zeros, so it's trailing off from wherever it was when the signal dropped under, though again this happens several times per wave cycle. Also, this is for hardware for sure, but some software works slightly differently. Mostly, it works out the same.
Wouldn't it be more correct to say "above 10Hz" instead of "but a pure square/sine wave"?
Click to expand...
I wrote what I meant, but I guess I was thinking about a full-wave rectified compressor in which a real square wave (with infinitely steep sides) would never actually cross the threshold and anything else will cross it four times per cycle. With half-wave rectification (pretty common in hardware if not in soft), even a square wave passes the threshold twice per cycle. Course, infinitely steep sides means infinite frequency response, which can't really happen without causing aliasing which should skew the results...

ReaComp makes a pretty flexible waveshaper/saturator/distorter. I don't use it as often as I might only because the curve of the knee makes it difficult to predict the actual output limit when ratio is set to infinity and knee at anything greater than 0.
 
ashcat_lt said:
I wrote what I meant, but I guess I was thinking about a full-wave rectified compressor in which a real square wave (with infinitely steep sides) would never actually cross the threshold
Click to expand...

I'd expect the input DC blocking cap to slope the sides and round the corners of the square wave so there would still be some signal, pointy spots where the slopes don't quite meet up (I say as someone who hasn't soldered in earnest for years).

Ooh, great name for a baseball team in our capital, the DC Blocking Caps.

OP, you getting all this?
 
bouldersoundguy said:
I'd expect the input DC blocking cap to slope the sides and round the corners of the square wave so there would still be some signal, pointy spots where the slopes don't quite meet up (I say as someone who hasn't soldered in earnest for years).
Click to expand...
You can't have infinite slopes in hardware. Not all software is as careful about AC coupling.

OP, you getting all this?
Click to expand...
;)
 
ashcat_lt said:
You can't have infinite slopes in hardware. Not all software is as careful about AC coupling.
Click to expand...

Hm, what happens when you generate a frequency above the Nyquist frequency of the sample rate you're using? I've actually done this before and the result is silence, but I don't recall if the signal simply isn't there (filtered out upon generation as an ADC would do) or if it's there but not audible (filtered out by the DAC). Pretty sure it's option #1.

Yep, option #1, filtered like an ADC, at least that's how it is in my ancient version of Sound Forge.
 
You must log in or register to reply here.
Back
Top