Clipping/distortion/lim/comp - what you do to your tunes

100 Replies, 19992 Views

that was distorting stuff by clipping it was it not, thats entirely different....

a compressor isn't compressing "constantly" it only kicks in when a specified threshold is reached....
beckett Wrote:that was distorting stuff by clipping it was it not, thats entirely different....

a compressor isn't compressing "constantly" it only kicks in when a specified threshold is reached....
No

i think this thread started talking about that, sort of Lol

the file macc posted on the first page was to show how badly done limiting/compression can introduce unpleasant distortion that might not be noticeable depending on how harsh it is but starts to get a bit nasty after you listened to it for a while

i think new york compression was talked about too at some point along with digital clipping Lol [Image: eekhands.gif]
beckett Wrote:it might make sense using it as a send effect on vocals perhaps....see Macc's "funk" thread, there are links in there that describe the "Excite" technique that sounds similar to the above scenario.
I was assuming he meant 'mixing uncompressed and compressed signal' ie something entirely different from what he actually wrote. Teef

I'd imagine that having compression as a send might make sense if you wanted a quite specific sort of compression on everything to varying degrees or something, but it's not something I'm in a rush to try out.

By the way, how did you manage to get a compressor as a send by mistake? On the desks I've used (for live stuff), that'd be like putting your ignition key in the CD player for a year without noticing - do your inserts not live in channel strips and sends somewhere over by the master outs?
tyler Wrote:but still, this whole idea contradicts maccers' tirade against people distorting stuff by compressing it?!
Icon_question
Bigstick
Wink
Didn't see this - bit bizzley but will read and have a look...

BTW there's nothing wrong with distorting parts through compression - jeebus, I love it. I get pissed off when people ruin their hard work by overlimiting and their bass causes everything in their tune to distort.

Just skimmed it and I am not sure what yer on about, tyler Baffled

What's contradicting what I am saying exactly? Baffled
[Image: protabl3.gif]
Don Cherry Wrote:Every human is blessed in her or his life with one love (passion), no matter how long it may last. This Absolute love will last in one's heart and soul forever.
Checked the samples tonight. I was shocked.
well from that sound you posted, i was thinking that getting distortion from your compression/limiting is a bad thing, because it doesnt sound so good, and one of your points was that if you limit too hard then even if you can't hear the distortion, it tires your ears.

so then when we started talking about compressing bass, i'd have said its a bad idea because of the nasty distortion mentioned above since you can just fix the dynamics in the synth, whereas people seem to be now saying that the distortion is a good thing because it adds harmonics Baffled

in terms of my ears i totally agree that the wav you posted sounds nasty, and you dont want to introduce that into your track, and that compressing bass makes it sound better than just changing the attack in your synth

but in theory i'm struggling to see how we've gone from this distortion sounding nasty to being something nice and harmonic we want to add Baffled

surely it can only be one of the two? i know at first we were talking about over limiting, but the example you posted was of compression, and you did say that you dont want it there even if its subtle because it tires the ears.. so in theory how can that not apply to compressing bass?

incidentally i think what is was banging on about convolution might be how the liquid mix works Chin

sorry if that comes across wrong just not following how all this fits together!
I thought that the key point of that example was that the unpleasant part only kicks in when the release gets really short, ie on the same order as the wavelength. Because then the compressor hears one peak in the waveform of the bass, starts reducing the volume, the peak finishes, the compressor thinks 'oh, that's good, start bringing the volume back up' and actually makes a significant change in the volume before the next waveform hits and it starts reducing the volume again.

Effectively (I think) you're modulating the amplitude with a triangle/sawtooth wave at double the frequency - because every positive or negative peak of the waveform triggers a little release period and thus a little bit of amplitude modulation. Obviously the precise shape of the modulating wave depends on your compressor.
slothrop's pretty much hit the nail on the head Xyxthumbs

perhaps not double the frequency exactly, but the point is that when the release time is of the order of the period (ie 1/f) you start to modulate the signal and not the gain.

or perhaps another way to put it is that you are modulating the gain of a single cycle (in whatever fashion), which is in effect changing the basic waveform. well, not even in effect - you are chaging the basic waveform.

good stuff peeps! :d Icon_yippee
[Image: protabl3.gif]
Don Cherry Wrote:Every human is blessed in her or his life with one love (passion), no matter how long it may last. This Absolute love will last in one's heart and soul forever.
Xyxthumbs
but then if you set the release long enough that it doesnt do that, how is it adding harmonics?
dionysus Wrote:Checked the samples tonight. I was shocked.

Try doing the same thing with L1 on the default release setting JEDIcrying

People do that to their entire tune!! I got yet another just yesterday!
[Image: protabl3.gif]
Don Cherry Wrote:Every human is blessed in her or his life with one love (passion), no matter how long it may last. This Absolute love will last in one's heart and soul forever.
tyler Wrote:but then if you set the release long enough that it doesnt do that, how is it adding harmonics?

There's a short answer and (probably extremely) long answer to this.

Short answer: it is still modifying the waveform.

I am drafting the long one in Word so it looks like I am working and that 'black and orange database' isn't on the screen the whole time Hahaha
[Image: protabl3.gif]
Don Cherry Wrote:Every human is blessed in her or his life with one love (passion), no matter how long it may last. This Absolute love will last in one's heart and soul forever.
this thread hurts my head.....what are u ppl's talking about????

Icon_cry
'There's no such thing as selling out just buying in'

Chuck D
Have you listened to the audio files?
[Image: protabl3.gif]
Don Cherry Wrote:Every human is blessed in her or his life with one love (passion), no matter how long it may last. This Absolute love will last in one's heart and soul forever.
Macc Wrote:CZHECKQUE IT

http://rapidshare.com/files/71945461/Re3...e.mp3.html

Guess who forgot to turn numlock off, hoho.

So look, this is a C0 sine wave being generated by the A1 synth, roughly 40 Hz I think it is.

It is put through Voxengo Marquis compressor - I did it with Waves RComp but the automation didn't work Roll Hahaha .

The compressor is set so that it is doing 10-11dB of gain reduction.

The file starts with the release at max - 8 seconds on Marquis. I turn the release down slowly and you can hear the expected distortion coming in, and it's no unpleasant on this source.


Then at about 36-38 seconds into the file, I get below 10 ms release. You hear the artifact I am on about at the top end. I then nudge the release back and forth a few times, going from no more than 15 ms, no less than 4ms. Listen to how quickly that effect fades in and out, even with such a small release time difference.


Again, listen to the high end. It doesn't sound like a sine wave any more, there's some raspy high end shit going on where the compressor is jumping up and down between the peaks of the waveform, modulating the audio itself. You can hear it rasping/drilling, not like a constant signal, or harmonic distortion.

The gain reduction meter also visibly jumps up and down when going below that point.

THIS IS WHAT YOU GET ALL OVER YOUR PRECIOUS TUNE WHEN YOU LIMIT BY 6dB WITH HEAVY BASS CONTENT AND TOO SHORT A RELEASE TIME.

Smile

now,I'm getting it.

Macc once again, coming with the science!
'There's no such thing as selling out just buying in'

Chuck D
macc Wrote:
tyler Wrote:but then if you set the release long enough that it doesnt do that, how is it adding harmonics?

there's a short answer and (probably extremely) long answer to this.

short answer: it is still modifying the waveform.

i am drafting the long one in word so it looks like i am working and that 'black and orange database' isn't on the screen the whole time Hahaha

black and orange database Lol

actually think i'm almost there, was forgetting about the release...

so when the release is longer, clearly it will still affect the waveform in some way Chin

basically is it that the extra frequencies you add with a short release are nasty, whereas when its longer the actual character of the distortion is different, not just the same nasty frequencies but quieter? i'd be down with that Teef although i'm still not 100% sure on why
got work come in so i have abandoned my draft Icon_sad Hahaha

but consider the following tyler;

on paper you might well be correct - to be frank i need to think about it a bit more. if it was as simple as you're suggesting, then all compressors would sound the same, and they don't.

so as you've intimated earlier in the thread, the key must be in the design, and interaction, of the sensing and (particularly) the level control components - the heart and soul of any compressor. i'll be perfectly honest and say that i don't know an awful lot about that part of compressor design though, so this is all semi-informed stream-of-consciousness stuff here...


it's fair to say that cubase sx's dynamics plugin doesn't sound the same as - to pick an example out of the air Teef - a hardware neve 33609. even when applying the same gain reduction with identical-if-not-then-as-close-as-possible settings, they sound markedly different. all compressors sound different. but in our imaginary 'sine-wave and long release' test they're all doing the same thing, right?

so it must be fair to say that the means by which gain reduction is achieved - even with a steady state signal - is just as important as the numerical reduction being applied. if it were a simple transfer curve (input value = x, draw a line up to the transfer curve, draw line across to get output value y) then what you have is simply a lookup table of input to output. it's a lot more complicated than that.

this is exactly why old (and i mean old) vst plugin compressors were derided - they were effectively nothing but a transfer curve. when you're using, say, a signal-dependent light element shining on a photo-sensitive resistor (each with their own particular behaviours) which is controlling the gain, it is a different story.


something that's really burning in my mind is how shit the first generation of antress modern plugins were. they all sounded the same and they all sounded like the level was being made constant and that was it. it sounded like a simple mathematical representation of the transfer curve. in hindsight i wish i'd done more testing (if just for this thread), but as the sound was so unfeasibly bland and unmusical i just deleted them (which says a hell of a lot in the context of this discussion!). i'm thinking that perhaps those plugins might demonstrate the behaviour you're driving at, if i could find the old versions again. that is;

if tr <1> or = 1/f then no buzzy shit, just gain reducxion (typo, but Hahaha ) and nothing else.


scouring the ua webzine (where they really lay out proper science - it's fucking great) brings up this from 2004;

http://www.uaudio.com/webzine/2004/february/index2.html

read that Xyxthumbs

notably;

Quote:compression has many different applications, and, as a result, many different methods of signal detection and gain computation have been developed. it is not surprising that the sonic character of a given compressor is heavily dependent upon the precise behavior of its detector and gain computer.

Quote:program-dependent release is a feature of many popular compressors. the motivation for having program-dependent release is as follows: after a transient, it is desirable to have a fast release to avoid prolonged dropouts. however, while in a continued state of heavy compression, it is better to have a longer release time to reduce the pumping and harmonic distortion caused by repetitive attack-release cycles.

and slightly less pertinently;

Quote:because of the many dimensions of behavior a compressor can exhibit, there is no cookie-cutter approach to modeling of analog compressors. each unit must be studied separately, and a separate model constructed. because of inherent nonlinearities associated with compression, there is no way to guarantee faithful behavior by signal modeling, or studying input/output behavior. only modeling of the physical processes happening within a given unit will result in a robust digital model.



Smile

looks like i should have written this in word Lol
Macc Wrote:If it were a simple transfer curve (input value = x, draw a line up to the transfer curve, draw line across to get output value y) then what you have is simply a lookup table of input to output. It's a lot more complicated than that.

This is exactly why old (and I mean old) VST plugin compressors were derided.

If Tr <1> or = 1/f then no buzzy shit, just gain reducxion (typo, but Hahaha ) and nothing else.


Lightbulbs going off all over the shop here;

This is exactly why the 'magic release time' trick works BETTER - or rather is much more clearly defined/audible - with shit VST plugins like mda limiter Icon123
[Image: protabl3.gif]
Don Cherry Wrote:Every human is blessed in her or his life with one love (passion), no matter how long it may last. This Absolute love will last in one's heart and soul forever.
my guess on the whole thing would be that the kind of shizzle you get in analog hardware stuff to detect peaks isnt perfect as it is in software, and has all sorts of frequency dependencies and stuff that gets passed on to the gain reduction

also the gain reduction may well have some non-uniform reduction going on across the spectrum in the way that it applies the reduction, and the attack/release envelope that is uses is probably very non-linear and changes shape depending on the input signal and all sorts of weird shit.

i tried thinking about all this in regards to just a shite software compressor and to be honest i'm confused
tyler Wrote:my guess on the whole thing would be that the kind of shizzle you get in analog hardware stuff to detect peaks isnt perfect as it is in software, and has all sorts of frequency dependencies and stuff that gets passed on to the gain reduction

Yes in the article i linked above there's a a mention of the frequency dependent attack and release characteristics of the la2a, for example.

Quote:also the gain reduction may well have some non-uniform reduction going on across the spectrum in the way that it applies the reduction

exactly - see above.

Quote:i tried thinking about all this in regards to just a shite software compressor and to be honest i'm confused

nah man - you're right with regards to a shit software compressor. in fact, your line of thinking is even more basic than the most basic of software comps. you are simply doing a look up table in your head (it seems to me anyway). but it depends what you are turning it down with innit Wink
[Image: protabl3.gif]
Don Cherry Wrote:Every human is blessed in her or his life with one love (passion), no matter how long it may last. This Absolute love will last in one's heart and soul forever.
Macc Wrote:In fact, your line of thinking is even MORE basic than the most basic of software comps. You are simply doing a look up table in your head (it seems to me anyway). But it depends what you are turning it down with innit Wink

i'm just thinking in terms of mr compressor notices its loud -> gain reduction

mr compressor starts an attack envelope towards his desired gain reducition (the gain reduction depends on how loud it is, ignoring that), then once he's there starts the release back to no gain reduction... then once thats back potentially does the same thing over

that might be wrong though, i'll have a gander at the musicdsp.org stuff later
Macc Wrote:So it must be fair to say that the means by which gain reduction is achieved - even with a steady state signal - is just as important as the numerical reduction being applied. If it were a simple transfer curve (input value = x, draw a line up to the transfer curve, draw line across to get output value y) then what you have is simply a lookup table of input to output. It's a lot more complicated than that.

This is exactly why old (and I mean old) VST plugin compressors were derided - they were effectively nothing but a transfer curve. When you're using, say, a signal-dependent light element shining on a photo-sensitive resistor (each with their own particular behaviours) which is controlling the gain, it is a different story.
Are we talking about better digital compressors not using a transfer curve here, or about them using a less linear transfer curve? Ie if a naive compressor has decided via its signal detection and its combination of threshold, ratio, attack and decay (and yeah, I guess that the way those interact is going to be another area where it'll be more naive and linear less sophisticated) that it needs to apply gain of 0.7 for a while, it takes the input stream and applies x -> 0.7*x and outputs that. Whereas your fancy-pants UAD compressor might apply something like x -> 0.68*x + 0.01*x^2 +0.01*log(1+9x) or something else that applies a little bit of very subtle waveshaping distortion.

Or is it something even weirder?

I'm beginning to see why people can charge so much for a good compressor. Teef
Different digital compressors use different transfer curves, you'd expect, but apart from providing different y values (ie output volume) this wouldn't by defintion affect the sound unless there are further consequences (when y = whatever then do such and such).

I don't see how it can make a difference at all to the sound when you're strictly simply reading a value from a graph. It's not like kids in GCSE maths are banging on about how warm their paper is Teef

It's the component (be it digital, analogue, modelled, whatever) doing the detection and the component doing the turning down and their associated behaviours and nonlinearities that does it. I'm not concerned about specifics here exactly - let's just say 'analogue' versus 'digital', both being at extremes of a spectrum (where exceedingly well-modelled stuff like UAD sits kind of in the middle).
[Image: protabl3.gif]
Don Cherry Wrote:Every human is blessed in her or his life with one love (passion), no matter how long it may last. This Absolute love will last in one's heart and soul forever.
Sorry, I was getting the transfer curve and the 'component doing the turning down' confused. I was basically thinking of a nonlinear amplifier.

Also, Casio graphic calculators were fucking incredible for getting the genuine 'paper' feel out of a digital graph.

Possibly Related Threads…
Thread Author Replies Views Last Post
  Clipping / Distortion Question mucsavage 14 4,686 8th December 2008, 18:38
Last Post: mucsavage
  analog clipping-good or bad?? Ruffneck Dave 6 2,942 25th February 2008, 14:47
Last Post: Ruffneck Dave
  master comp phoonkah 31 12,656 16th January 2008, 15:45
Last Post: subvert
  Oldskool Drum Distortion using Cassette tapes! bhima 12 3,321 5th October 2007, 15:35
Last Post: SETI
  What comp to use for sidechaining? markgabba 13 4,747 29th July 2007, 20:14
Last Post: Macc