How is gain reduction applied by compression?

Incanus

Incanus

Nerd-man
The outboard gear I'm working with is an RNC and a JoeMeek twin-q. I'll probably occasionally use a plug-in from time to time as well.

But my question is:

How does a compressor apply gain reduction to a given sound? Specifically what I want to know is: does it always attenuate the highest peak in the signal, or are there certain frequencies that a compressor goes after first, like highs or lows? So, what if a signal had no peaks at all, if it were a 'flat' sound (probably nigh impossible, but for theory's sake. . .)? Would the compressor go after the low end or the high end?

Or, am I just a complete dolt for asking such questions? It's quite possible. I would just like to understand how compression is applied to a signal a little better. Any info into this matter would be greatly appreciated.
 
on a compressor there is a knob called "threshold". Anything that goes higher than this threshold is compressed with the settings you set. most compressors will show you the input level of the signal as well as how much you are reducing...this way you can decide how much of the signal you want compressed.
 
"Gain reduction" just refers to how strongly the compressor dampens the signal. a 3:1 ratio of reduction, for example, means that for every increase in volume of 3dB that is going into the compressor, there is only going to be a 1dB increase in volume coming out of the compressor. a 4:1 compression ratio for gain reduction means a 4dB increase going in will be only a one dB increase going out. And so on.

On standard compressors this gain reduction is applied to all peaks, regardless of frequency, as long as the peak volume level exceeds the "Threshold" setting. A threshold setting of -8dB would mean that any signals peaking below -8dB at the input will pass through the compressor untouched; only those peaking louder than -8dB at the input will be acted upon with the gain reduction ratio.

There are "multiband compressors" in which you can apply seperate compression settings to specific frequency ranges, often low, mid and high frequencies, but other than that they work just like normal compressors. They are kind of like 3- or 4- band equalizers, but instead of applying volume to a certain frequency range like an EQ does, they apply compression to that frequency range instead.

HTH,

G.
 
Thanks for the replies.

I think I got it now. I was over-thinking this.
To put it simply, every and all peaks above the threshold gets compressed. So in my 'flat' sound example above, the entire sound would get compressed because all peaks would have crossed the threshold.

And that explains why I lost some low-end on a snare drum I had recorded: the low end was the only thing crossing the threshold!

I feel a little stupid for having asked this question now, but I did learn something, so what the hell.
 
Incanus said:
Thanks for the replies.

I think I got it now. I was over-thinking this.
To put it simply, every and all peaks above the threshold gets compressed. So in my 'flat' sound example above, the entire sound would get compressed because all peaks would have crossed the threshold.
Click to expand...
If by "flat" you mean basically a steady DC signal, that at it's outset rose above the threshold and then stayed there, then yes, compression would have been applied at the point of attack (assuming that the rate of attack and the attack speed setting on the compressor jived, but that is another topic which we've kind of ignored up until this point.)

Incanus said:
And that explains why I lost some low-end on a snare drum I had recorded: the low end was the only thing crossing the threshold!
Click to expand...
That could be. It could also be not a threshold issue, but an attack issue. If the attack on the compressor was set too slow, it's possible that the high freq transients caused by the actual striking of the drum skin "got past" the compressor because it couldn't react fast enough to them.

Incanus said:
I feel a little stupid for having asked this question now, but I did learn something, so what the hell.
Click to expand...
Oh stop that! :) There are no stupid questions except those that start "walters" threads. ;)

G.
 
Thanks Glen.

What I meant about the 'flat' signal was for theoretical purposes as there probably is no such thing in reality. My theoretical sound had the exact same amount of db at every frequency. So once it crossed the threshold, it would ALL be compressed.

I should research more, but is it true, then, that high frequencies travel faster than low frequencies? That would make the attack setting quite crucial in this regard.

APL, I'll need more time than my lunch break to check out the article you suggested. Later tonight, or tomorrow, or something. Looking forward to it. Thanks!
 
Incanus said:
What I meant about the 'flat' signal was for theoretical purposes as there probably is no such thing in reality. My theoretical sound had the exact same amount of db at every frequency. So once it crossed the threshold, it would ALL be compressed.
Click to expand...
Ah, yes, I ms-interpreted what you meant by "flat". By your description, yes, the compressor doesn't care what frequency the signal the signal is oscillating at; as long as the shape and size (attack speed and threshold) of the signal match the compressors settings it will act upon it.

Incanus said:
but is it true, then, that high frequencies travel faster than low frequencies? That would make the attack setting quite crucial in this regard.
Click to expand...
Weeelllll...that is kind of a loaded question when you really dive into technicalities and physics and such. For the basic parameters of this discussion, let's keep it simple for now and say that attack speed and freqency are *not necessarily* directly interrelated. It is possible to have very fast low-frequency attacks; for (extreme) example a hard slap to an electric bass will probably have a faster attack than a drummer using brushes on his hi hat.

The reason I mentioned that was because you mentioned your snare situation. Often times there, the fastest atack is the sound of the stick hitting the skin, which happens to be at a higher frequency than the resonant sound of the snare drum itself, so in that cas it could have indeed been that the high freqs blew past the compressor. But I think to turn that into a general rule here is a bit much.

Hopefully I haven't confused you more now than I did when we started. :D

G.
 
SouthSIDE,

I really appreciate your helping me think this through.

OK. Here's my attempt at paraphrasing and generalizing what you were last speaking of: It is really a matter of understanding the characteristics of the sound you are recording so that you can apply the proper attack and release times along with the threshold and ratio to compress the sound the way you want it (the way it sounds best).

Hmmm. I tried. . .
It's no less confusing than what you wrote, and maybe more so!

I still have to read apl's link, too.
 
SouthSIDE Glen said:
"Gain reduction" just refers to how strongly the compressor dampens the signal. a 3:1 ratio of reduction, for example, means that for every increase in volume of 3dB that is going into the compressor, there is only going to be a 1dB increase in volume coming out of the compressor. a 4:1 compression ratio for gain reduction means a 4dB increase going in will be only a one dB increase going out. And so on.

On standard compressors this gain reduction is applied to all peaks, regardless of frequency, as long as the peak volume level exceeds the "Threshold" setting. A threshold setting of -8dB would mean that any signals peaking below -8dB at the input will pass through the compressor untouched; only those peaking louder than -8dB at the input will be acted upon with the gain reduction ratio.

There are "multiband compressors" in which you can apply seperate compression settings to specific frequency ranges, often low, mid and high frequencies, but other than that they work just like normal compressors. They are kind of like 3- or 4- band equalizers, but instead of applying volume to a certain frequency range like an EQ does, they apply compression to that frequency range instead.

HTH,

G.
Click to expand...

Thanks for that, a clear explanation.

Can i guess that a basic rule of thumb might be that, say, if you want a signal around the -6db range, and your current signal is peaking at -2db, you might start with threshold at -8db and set the gain reduction to 3:1 (the idea that you are taking your peak at -2db, and then compressing the difference between -2db and -8db (6db) by 3:1 for a 2 db peak beyond -8db or -6 db.

Or does it not work out that simply and end up killing all dynamics when you are done or something?
 
daav said:
Thanks for that, a clear explanation.

Can i guess that a basic rule of thumb might be that, say, if you want a signal around the -6db range, and your current signal is peaking at -2db, you might start with threshold at -8db and set the gain reduction to 3:1 (the idea that you are taking your peak at -2db, and then compressing the difference between -2db and -8db (6db) by 3:1 for a 2 db peak beyond -8db or -6 db.

Or does it not work out that simply and end up killing all dynamics when you are done or something?
Click to expand...


I can't say that I entirely understand your question. But I think I can safely say that 'Gain Reduction' is not synonymous with 'Ratio'. The same Ratio (i.e. 3:1) can result in different amounts of gain reduction depending upon the other settings of the compressor (particularly the Threshold). If you are trying to get 6db of GR, go by your GR meter, not the ratio.
 
Incanus said:
I can't say that I entirely understand your question. But I think I can safely say that 'Gain Reduction' is not synonymous with 'Ratio'. The same Ratio (i.e. 3:1) can result in different amounts of gain reduction depending upon the other settings of the compressor (particularly the Threshold). If you are trying to get 6db of GR, go by your GR meter, not the ratio.
Click to expand...

Well that is pretty close to addressing what i was asking. I guess ia m trying to go down a hypothetical road, so please don't take the idea that i want to rtemove the ear or trail and error with tweakingout of the picture. But my question boils down to:
Is it possible to take a target db level (-6db say) and the pre-compressed peaks (-2db in my example) and "eyeball" the threshold and ratio (I agree that depending ont threshold and general dynamic level of the signal, differrent amounts of dynamic changes should happen, but I do think that everyhign have read indicates this is a true ratio).

So anyway, if i am shooting for the -6db mark, and am 4 db above it, would it make sense to estimate a threshold below the mark, taking into account my peak of all peaks and the 3:1 ratio, intending to hit the -6db mark or so.

I know, I know, the real answer is to sit down and play with it to get the real feel for these setting, but the excercise is interesting to me.

Thanks anyway :)
 
IME the RNC is kind of a Magic Box used in Super Nice mode, which is the only mode I use it in. (I use the RNC only for unamplified gtr.) It levels the signal based on its response to the broadband energy in the signal - with guitar usually triggered by low mids. I like it because it levels in a way that doesn't create the usual high freq compression artifacts.

A multiband comp does it differently, where each freq band senses level and compresses it however you set it. But the real issue with multiband comps, IMO, is the sound quality of the filters that divide up the freq spectrum. Most of them, to my ear, cause horrible problems with the sound that make them unusable - totally separate from any actual compression that's used. The Waves Linear multiband comp is good IME, as is the Slim Slow Slider C3, which is freeware.

Tim
 
Last edited:
Incanus, probbly the easiest way to grasp what the compressor is doing, is to smash the shit out of something, and listen to how terrible it sounds with all the frequencies and overall sound so closely matched. Takes all the dynamics out of it, and music is all about the dynamics. Then, start backing off some of the settings, like ratio, and leave everything else the same. You then get a feel for how the amount of ratio affects what you are listening too. Then. try it again, but bring up the db. See how different that is. I'm by no means an expert on this, but have played with mine a bit this way, and heard how it affaects things. It's a very interesting tool. I've made some guitar tracks that look very flat on the pc screen. They sounded weird, but were interesting.
You can kill a good drums sound just by attacking too quick, and releasing very slowly. But on some shitty vocals, like mine, this procedure might help, at least a bit. Different sounds, different techniques. Read some general ideas, play with them, and change them to your liking. Some of what you do may not need anything like what I, or anyone else needs. I've heard some vocals that are awesome, without any compression. Mine, well they should be silenced, but lots of compression makes them bearable.
Ed
 
Oh yes, I tweak around for sure. And that is the best way to learn. But I am looking to bring a little more knowledge into how I go about compressing the various instruments I record.

My ultimate goal, if it is indeed possible, is to understand compression well enough to sort of counteract the effect it has on the sound of the instrument being recorded/mixed. To my ears, even a tiny amount of compression in Super Nice mode (on an RNC) changes the sound of the instrument at least a little. I find the effect of taming the dynamics of certain tracks slightly to be very benificial, but would like to keep as much of the original sound as possible. I can always tweak and use it to drastic effect when I need to, but would like to be able to make it as invisible as possible when I need to do that too.

Or am I just out of my mind to have such a goal? It's plausible.

But thanks, everyone. I've already learned a-couple-a more things than I new before I started this thread.

This place rules!
 
You must log in or register to reply here.
Back
Top