Different in noise. Can someone explain pls?

[Autobot]

I'm aware of the different colors of nice like brown, pink, white, etc. noise but what exactly is Gaussian, F-F2-1/F2, Crackle, Dust, Bow, MM5837, LFClipNoise, Logistic, Lorenz-Rossler, Wave, etc. noise?

I found a Max4Live device (http://maxforlive.com/library/device/4588/fp-noise) with those different types of noise and wonder whats about. Can those type of noises reproduce with a synth like Zebra or similar.

Someone here to share knowledge? Thanks.

[BertKoor]

I'm aware of the different colors of nice like brown, pink, white, etc. noise but what exactly is Gaussian, F-F2-1/F2, Crackle, Dust, Bow, MM5837, LFClipNoise, Logistic, Lorenz-Rossler, Wave, etc. noise?

I found a Max4Live device (http://maxforlive.com/library/device/4588/fp-noise) with those different types of noise and wonder whats about.

You could do some googling and find:
https://en.wikipedia.org/wiki/Gaussian_noise
https://www.musicdsp.org/en/latest/Synt ... ators.html
http://synton.nl/wp-content/uploads/2017/04/mm5837.pdf

So most of these types are about the calculational model used to generate it. Just run it and take a look at the spectrum it generates.

Can those type of noises reproduce with a synth like Zebra or similar.

The Noise module features four different flavours: White, Pink, Digital, and Crackles - which produces sounds simlar to a Geiger counter

So that's "no, not directly"

[Autobot]

Thanks. I used google but since I'm not a scientist the explanations did not brought me further I would be more interested in the sound of the different noises. As I said I'm familiar with different colors of noise and their sound but how does sound MM5837 noise? As far as I understand the MM5837 is a chip for generating pink/white noise. So whats the difference in terms of sound?

Edit: maybe I buy the m4l device to test the sounds by myself. although for the 3€ I could buy groceries for a little dinner. Yes! I'm tight on money...

[Autobot]

sorry. double post.

[andro_b]

Take a look at Dr. Stéphane Pigeon's page of audio tests, in particular here:

https://www.audiocheck.net/testtones_index.php

He gives some subjective impressions and perceptions of what different 'colours' of noise sound like, and an explanation of each term.

It is indeed a mathematical subject, but the above has good verbal descriptions, and sound samples.

Things like crackle and dust are not mathematical noises, but usually based on csound generators which model those phenomena, using algorithms to emulate such things. Various synths have varying numbers of different types of noise, but usually white and pink at a minumum. If you want lots of noise generators, the software (and hardware) modular synths have hundreds of noise modules available.

[mystran]

Things like crackle and dust are not mathematical noises, but usually based on csound generators which model those phenomena, using algorithms to emulate such things.

It is also possible to imlement such thing as simple samples, looped and started from a random position (just make them long enough that the repeat isn't obvious), or you could even do some sort of granular synthesis for them. This way you can easily have "noise generators" that produce sounds that are tricky and/or expensive to model properly.

Another classic approach to generating "standard" noise with arbitrary spectrum is to take a long FFT window (eg. at least 10-15 seconds), fill the magnitudes with whatever spectrum you want, randomize the phases and then IFFT the thing to get a perfectly looped sample.

[Music Engineer]

if you read something like "gaussian", this refers to the probability density distribution of the noise - that is the relative number of samples that fall in a given range of amplitude values. the most basic noise generators will give you a uniform distribution of the amplitude values. to see this, you want to look at the histogram. for uniform noise, it looks like this:

NoiseUniform.png

every amplitude value is equally likely to occur. if you add the output of two such uniform noise generators (and divide by two), you get a triangular distribution like this:

NoiseTriangular.png

and if you add the output of three (and divide by 3), you get this:

NoiseQuadratic.png

this is actually a piecewise parabolic curve. if you add n noise generators, the theoretic shape is the convolution of n rectangular functions which gives you a piecewise (n-1)-th order polynomial. the next one would be piecewise cubic, etc. and the shapes become more and more gaussian'ish, the more noise-generators you add. this is called the Irwin-Hall distribution. note that all these noises are spectrally white - and also sound more or less the same to me (although, of course, the uniform is loudest because it has the highest amount of high-amplitude values) but the distribution of amplitude values may likely have effects on nonlinear processors further down the signal chain (just a hunch - i did not yet explore this very much).

[Music Engineer]

just a few days ago, i played around a little bit with combining basic noise generators. here is an example of some sort of crackly noise (i needed to zip it because the forum does not support wavefile attachments):

NoiseTriModal7.zip

what i did here was the following: use three irwin-hall generators (of order 7) with different centers (one in the middle, i.e. mean=0, one with negative and one with positive mean), use another noise generator to select, per sample, which one of the three to use for the current sample. the selector generator is lowpassed in order to make it more likely to select the same from the 3 for successive samples. the result is some sort of crackle. this noise has the following distribution:

NoiseTriModal.png

it has 3 maxima, which is why i called it trimodal. but the histogram does not tell you the full story - i could create noises with the same distribution without the lowpass on the selector but they would sound totally different - the crackle would be gone. note also that the frequency spectrum also doesn't really tell you very much here - it looks almost white, too with a slight boost of lower frequencies. i wonder which sort of plot would reveal the features of this sort of noise. maybe some sort of histogram that shows the conditional probability of amplitude values, conditioned on what the previous value was - or something. i guess, if one likes noises, one can have a lot of fun by combining noise generators in various ways

[Music Engineer]

lorenz and rössler systems are a totally different thing. they are based on chaotic systems of differential equations:

https://en.wikipedia.org/wiki/Lorenz_system
https://en.wikipedia.org/wiki/R%C3%B6ssler_attractor

lorenz sounds like this:

LorenzX.zip

maybe not totally useless. physics is full of such chaotic systems. might be fun to explore. i didn't do rössler yet, but i can serve with double-pendulum instead:

DoublePendulumM1_2.zip

[BlitBit]

every amplitude value is equally likely to occur. if you add the output of two such uniform noise generators (and divide by two), you get a triangular distribution like this:

NoiseTriangular.png

First of all, thanks for this great visual explanation! Do I understand correctly that this basically means: If you take two white noise sources and mix and normalize them then it will be more likely that two mixed samples will cancel out than that they result in a "large" value?

[Music Engineer]

If you take two white noise sources and mix and normalize them then it will be more likely that two mixed samples will cancel out than that they result in a "large" value?

yes, indeed.

[earlevel]

First of all, thanks for this great visual explanation! Do I understand correctly that this basically means: If you take two white noise sources and mix and normalize them then it will be more likely that two mixed samples will cancel out than that they result in a "large" value?

If it helps to think of natural things, consider that int there were only a single source of noise in electronics, it would be random, and sound like white noise. But electronic devices have many noise sources 9and we tend to further mix multiple electronics devices together as well, in audio)—it still sounds white, but the probability distribution is Gaussian.

Summing only two sources of completely random, you get a triangular distribution. Again, it still sounds white, but while a single random source is just as likely to be at any amplitude, a triangular distribution will spend more of its time in the middle. It's the same reason we use a pair of dice in craps. There is only one way to roll 2 (1+1) or 12 (6+6). But there are six roll 7 (1+6, 2+5, 3+4, 4+3, 5+2, 6+1). Plot out all the possibilities, you'll see the triangle (five ways to roll 6 or 8, four for 5 or 9...). Most often, we want bipolar noise, so the "middle" is 0 and the position and negative peaks are hit less often.

There's a nice graphic of this in my video here (the link is queued to that part if you click it, but otherwise you can forward to 1:53 to see the pretty dice I spent some time on), if you want to see why someone would want triangular (Gaussian would also work for dither, it just requires summing more sources, while triangular is adequate for the job). Audio Dither Explained