Compression and saturation, in what order?

[Music Engineer]

when you do really heavy saturation - that is, in fact distortion - then your output-signal will have hardly any dynamics anymore. so no need for compression here anymore. the other way around: as your saturator does take away the dynamics from the signal anyway, there's no need to compress it beforehand.

edit: but the second case makes nevertheless more sense to me. you can then see the compresoor as a device which ensures, that you drive your saturator always in approximately the same region

[tahome]

I beg to differ. A compressor smoothly equalizes the changes in the dynamics of a signal. Saturation on the other hand just chops off everything that is above a threshold.

The difference of that is that saturation actually ADDS something to your signal (distortion) while a compressor just smoothes out the volume changes.

IOW: saturating a signal means trading high levels for distortion while keeping the relationship between soft and loud passages approximately the same (note that clipping does NOT really reduce the signal's power, but it does reduce its peak level!), while a dynamics compressor changes the ratio between the volumes of soft and loud passages, ideally without adding distortion (a compressor reduces both peak and power levels).

In fact, both treatments can make sense, even if used at the same time.

--th

[Music Engineer]

didn't want to state, that it doesn't make any sense to use these two processes at the same time. i just wanted to point out, that distortion leaves not much of the original signal dynamics. heavy distortion makes the amplitude envelope essentially flat.

[#The fREaK!]

If you use smooth compression with long attack to give the sound a certain character, you can apply saturation afterwards to relatively gently reduce the overshoots, while applying the effects in the opposite order would soften (and "warm" and/or audibly distort) the sound more and then make the compression more transparent. Using harder compression in order to flatten the dynamics, you get (apart from that the compression would still be more transparent after saturation) the opposite result from the placement of the saturator.

[tahome]

didn't want to state, that it doesn't make any sense to use these two processes at the same time. i just wanted to point out, that distortion leaves not much of the original signal dynamics. heavy distortion makes the amplitude envelope essentially flat.

Yes, by chopping off the peaks you flatten the amplitude envelope. But this does NOT really change the dynamics of the signal much because the *power* of the signal is still at approximately the original dynamics. A compressor will smooth out both power and peak amplitudes. Both processes are related but different.

--th

[pekadan]

Cool! I really learned a lot from this discussion. Thanks all!

[Music Engineer]

didn't want to state, that it doesn't make any sense to use these two processes at the same time. i just wanted to point out, that distortion leaves not much of the original signal dynamics. heavy distortion makes the amplitude envelope essentially flat.

Yes, by chopping off the peaks you flatten the amplitude envelope. But this does NOT really change the dynamics of the signal much because the *power* of the signal is still at approximately the original dynamics. A compressor will smooth out both power and peak amplitudes. Both processes are related but different.

--th

maybe we are talking a little bit on cross-topics here - i'm talking about really heavy distortion. taken to the extreme, this turns everything into something like a square-wave. which has flat amplitude envelope and a crest-factor of 1.0

[tahome]

maybe we are talking a little bit on cross-topics here - i'm talking about really heavy distortion. taken to the extreme, this turns everything into something like a square-wave. which has flat amplitude envelope and a crest-factor of 1.0

For a pure input wave, yes it does.

But if you have a realworld input signal (such as speech) saturation will only approximate a square wave (the peaks are chopped-off) but it will still have variations in its dynamics (because it is a realworld signal that isn't of constant amplitude).

Chopping off the peaks does indeed limit the peak amplitude, but the area (= power) of the curve still isn't limited by this process. So the power of the signal changes in (approximate) proportion with the original dynamics while the peak values are nailed to the min and max of the available range. This does not really compress the dynamic range (unless, of course, you use ridiculously high values for the saturation. But in that case you could use a square wave to begin with and don't have to worry about your input signal ).

If you use a compressor (which *scales* the peaks to be inside the range as opposed to chopping them off) this is different - a compressor alters both power and peak values.

--th

[bobby yarrow]

No, not true. You might have misunderstood what I was referring to.

If you have a signal chain like this

Sound -> Reverb -> Delay -> EQ -> Output

then the net effect is the same as if you had

Sound -> EQ -> Delay -> Reverb -> Output

or any other combination.

That's just incorrect.

Here's a track with reverb>delay: www.theyarrows.com/temp/reverbdelay.mp3

Same track, same effects, delay>reverb: www.theyarrows.com/temp/delayreverb.mp3

Here's both tracks, with one out of phase: www.theyarrows.com/temp/phasedifference.mp3

I can't imagine what you mean.

[tahome]

If you route it like I said there is no difference. This is not something that is open for debate, it's a fact (read any book on DSP if you don't believe me).

Here's a link to get you started: http://www-ccrma.stanford.edu/~jos/filt ... ition.html

It says:

"When two signals are added together and fed to the filter, the filter output is the same as if one had put each signal through the filter separately and then added the outputs (the superposition property)."

And as you know, for addition the order doesn't matter. a+b is the same as b+a.

The difference between your tracks might come from the fact that the reverb uses modulated delay lines - in that case the result will be different on each pass and you cannot simply subtract the two files. That's why I said "the net effect" will be equal. It will always be equal if you reset the effects/modulation before doing the subtraction.

--th

[Jason Brian Merrill]

ummm..

there isnt any difference if you have, say a severly distorted guitar track nd you compress it after the fact.

[tahome]

@Jason: Yes there is a difference. Distortion and compression are both non-linear effects, in which case their order in the effect chain is important.

To reiterate what I have said about linearity, here's an example from a DSP book if you don't believe me:

http://www.dspguide.com/CH5.PDF

Check out page 89 which says that additivity is a requirement for a linear system. Check out page 90 for a definition of additivity.

HTH,
--th

[Krakatau]

If you route it like I said there is no difference. This is not something that is open for debate, it's a fact (read any book on DSP if you don't believe me).

Here's a link to get you started: http://www-ccrma.stanford.edu/~jos/filt ... ition.html

It says:

"When two signals are added together and fed to the filter, the filter output is the same as if one had put each signal through the filter separately and then added the outputs (the superposition property)."

And as you know, for addition the order doesn't matter. a+b is the same as b+a.

The difference between your tracks might come from the fact that the reverb uses modulated delay lines - in that case the result will be different on each pass and you cannot simply subtract the two files. That's why I said "the net effect" will be equal. It will always be equal if you reset the effects/modulation before doing the subtraction.

--th

I question that suddently comes to my mind : in the case a reverb uses somes dumping factors...
( for, as example, hi-mid-low frequency dampers that i assume they are somehow similar to a filtering process inside a feedback loop but more elaborate )
...i feel that it should be different if you use an EQ PRE or POST reverb


....so ( no sarcasm, trust me freely, if ever.. ) i'm just wondering in the case, where is the non-linear process ?

[Jason Brian Merrill]

@Jason: Yes there is a difference. Distortion and compression are both non-linear effects, in which case their order in the effect chain is important.

To reiterate what I have said about linearity, here's an example from a DSP book if you don't believe me:

http://www.dspguide.com/CH5.PDF

Check out page 89 which says that additivity is a requirement for a linear system. Check out page 90 for a definition of additivity.

HTH,
--th

screw textbooks, the sound tells me it doesnt do anything. perhaps a bit of pumping inbetween silent spots -- but not otherwise.

[tahome]

I question that suddently comes to my mind : in the case a reverb uses somes dumping factors...
( for, as example, hi-mid-low frequency dampers that i assume they are somehow similar to a filtering process inside a feedback loop but more elaborate )
...i feel that it should be different if you use an EQ PRE or POST reverb


....so ( no sarcasm, trust me freely, if ever.. ) i'm just wondering in the case, where is the non-linear process ?

There is none. As long as there isn't a non-linearity (like distortion, compression, modulation) the result will be the same regardless of where your EQ is with respect to the reverb.

If you put a damping filter into the feedback loop that would still be a linear process, but you would force the filter to be a recursive (IIR) one, while in the other case (putting it at the output after the reverb) it needn't be recursive.

IOW: it's a different filter. Still linear, but different.

--th