Talk:Colors of noise

Warning: Some of these seem to originate exclusively from a single newsgroup posting from October 1996 by Joseph S. Wisniewski, and should be considered questionable. The nature of the wiki will cause the definitions to converge towards the true definitions with time, making this article more accurate than the copies of that posting. Providing references for each will help with this.

Measurement 'noise'

This article seems to focus only on noise in terms of sound and maybe electronics. Where should something be included about noise (as a source of error) in making measurements - whether from technical or experimental source material? Of course, the same power/frequency concepts apply. Pat Heslop-Harrison 07:43, 21 May 2006 (UTC)

Real definitions

we could expand all of these to use the 1/f^β notation, where β = something, as used in some of these examples. — Omegatron

"The Federal Standard 1037C Telecommunications: Glossary of Telecommunication Terms defines four noise colors (white, pink, blue & black) and is considered the official source."

An official source, perhaps but any one nation's standards can't be taken to be the standard. Is there any ISO standard, for example? I'd have thought a lot of this information is worth putting onto the main page, BTW. --Douglas 4 July 2005 15:17 (UTC)

It's better than a newsgroup posting. Here are the official definitions:

white noise

Noise having a frequency spectrum that is continuous and uniform over a specified frequency band. (188) Note: White noise has equal power per hertz over the specified frequency band. Synonym additive white gaussian noise. [1]

pink noise

In acoustics, noise in which there is equal power per octave. [2]

black noise

Noise that has a frequency spectrum of predominately zero power level over all frequencies except for a few narrow bands or spikes. Note: An example of black noise in a facsimile transmission system is the spectrum that might be obtained when scanning a black area in which there are a few random white spots. Thus, in the time domain, a few random pulses occur while scanning. [3]

blue noise

In a spectrum of frequencies, a region in which the spectral density, i.e. , power per hertz, is proportional to the frequency. [4]

— Omegatron 01:45, 1 December 2005 (UTC)

Maybe it should read "Purple noise" because "Orange noise" isn't listed below. I don't know. I guess it's a matter of precision that I don't know about. Peter 03:51, 4 December 2006 (UTC)

Red

red = brown (1/f²) definition: [5] [6]

A mean reverting stochastic process http://riskinstitute.ch/00012612.htm

A stationary and causal Gaussian first order autoregressive (AR(1)) process {Xt} with mean zero (a real oceanographic definition) http://faculty.washington.edu/dbp/PDFFILES/red-noise.pdf

Geophysical processes, for example, are often characterized by red noise backgrounds. For red noise, the variance decreases with increasing frequency. In some cases, the overall noise trend can be approximated by a first order AR model. http://www.systat.com/products/TableCurve2D/help/?sec=1080

http://www.atmos.ucla.edu/~csi/REF/pdfs/ensomjo.pdf

I have a feeling red and brown are actually the same, and red/brown is used as a model for oceanographic research. In other words, the ocean-filtered noise is not red, but is modeled by red? - Omegatron 15:54, Jun 6, 2005 (UTC)

Why put up examples that can't be played with Quicktime, Real or MS?

Using Ogg Vorbis files is Wikipedia policy. There's a handy help link next to each one that will tell you how to configure your player to play them. Wikipedia uses Vorbis because it's not patent encumbered, the standard is totally open, and therefore it is the most likely to a) be usable far into the future when current formats die out, b)have a player for virtually any platform a user might be using, and c) isn't patent encumbered and thus Wikimedia can legally use any free encoders that exist --Ktims 00:56, 28 March 2006 (UTC)

Black

Most environmental noise is reddened: the variation is dominated by long-term £uctuations. Recent modelling has shown that moderately reddened noise a¡ects populations di¡erently from the white noise used in earlier studies. However, some geophysical phenomena, such as temperature and river height, can have deeply reddened `brown' or even `black' spectra http://oak.cats.ohiou.edu/~cuddingt/pubs/proceedings99.pdf

Green

"Bounded Brownian noise" - http://www.dxarts.washington.edu/courses/565/clm-2/green.cl

"the mid-frequency component of white noise" - http://www.engr.uky.edu/~dllau/Halftone/HtmlFiles/paper2.html Used in halftone dithering

More definitions in the footnotes here: [7] — Omegatron (talk) 15:58, 12 April 2008 (UTC)

Green noise

So why is there supposedly a 500 Hz hump in "green noise"? Rotation of the ether? --Chinasaur 00:50, 16 Jun 2005 (UTC)

Who knows? Original newsgroup post said "Green noise (defined by some folks producing relaxation tapes, Mystic Moods, I believe)". I don't know why this was removed. - Omegatron 01:36, Jun 16, 2005 (UTC)

Grey noise sample?

Great article, folks :-) Is anybody working on a sample of grey noise to complement the others - cause otherwise I might try to create one. Peter S. 10:20, 26 September 2005 (UTC)

Huh. I was going to do that. Must have forgotten. Feel free. I was just going to take white noise and FFT filter it in cool edit/audition, plugging in values from the weighting filter a-weighting equation. There's probably a better way. — Omegatron 13:51, 26 September 2005 (UTC)

No, just go ahead, I would have to search for an ogg encoder first, you seem to be already set up. Cheers :-) Peter S. 14:37, 26 September 2005 (UTC)

Alright, I'll try to remember to do it tonight. — Omegatron 16:39, 26 September 2005 (UTC)

Cool, many happy thanks :-) Peter S. 17:18, 26 September 2005 (UTC)

Oh, I remember why I never did this. Since it's the inverse of a weighting filter, and a weighting filter tends towards −∞ dB, the signal will tend towards +∞ dB at the frequency extremes. So I didn't really know how to weight it and still have a decent signal level. If I chose the range from 10 Hz to 20 kHz, for instance, the level at 10 Hz would be 0 dB, and the level at 1 kHz would be -70 dB. — Omegatron 21:02, 26 September 2005 (UTC)

Good point. (EDIT) I'd create a version of 20Hz-20kHz@16bit (19Hz would be the lowest frequency great cinema subwoofers output). At this range, you've got 55 dB between 20 Hz and 1kHz, plenty enough in 16bit space (~96dB). 24bit would give you a better s/n ratio, of course, but is it really necessary to have a low noise floor on a sound file that is essentially noise? :-) And in any case, add that disclaimer to the article. At least that's what I would do. Cheers! Peter S. 23:20, 26 September 2005 (UTC)

Oh, here it is. Sounds good, many thanks :-) Peter S. 09:39, 27 September 2005 (UTC)

Yeah, these are just approximations. I guess it all gets lost in the ogg conversion anyway, so it doesn't matter, but I was thinking of making it perfect, taking real random numbers from random.org, converting to sound file (I got this far last night before realizing it wasn't really worth it), making gaussian distributed, save as white noise, integrate in matlab or something (instead of FFT filter approximation), normalize, save as brown noise, etc. etc. If we could upload WAV files as well as listenable files I could see how this could be useful for posterity kind of reasons; do everything exactly so other people can use the files for other purposes, but for now it's silly overkill. — Omegatron 14:04, 27 September 2005 (UTC)

We can indeed upload lossless files, I have discovered, by encoding as ogg FLAC. I might go through and do it again sometime with perfect random numbers and mathematical accuracy. But not today. I'm not sure what it would accomplish, even, but it would be interesting to do... — Omegatron 00:45, 19 June 2006 (UTC)

Too complex

I'm not a physicist and understand almost none of this, could some of it be explained or simplified somehow? Fantom 17:36, 27 November 2005 (UTC)

Well, maybe we could add a graph of the audio spectrum for all the "noise colors" to make it easier to understand? Peter S. 18:36, 27 November 2005 (UTC)

Pictures would be very good. — Omegatron 00:53, 1 December 2005 (UTC)

I've added some simple spectrum analyses to the page, but the only source data I have is the .ogg files, so it's probably severely distorted from what a real graph would look like. I don't have the software handy (or the background knowledge/maths) to generate new files for some of the more exotic noises. If I get the .wav files, I'll improve them. Aside from Brown and perhaps Grey though, they look pretty good. --Ktims 14:27, 13 March 2006 (UTC)

Those are very good. I was going to make some, but yours are better.

1. I have the original wav files, and can send them to you if you'd like (send me an email). The spectra won't be much different except at high frequencies, but I like accuracy, too.
2. I'd like them better if they were log-frequency dB-amplitude, to show the "straight lineness" of the spectra.
3. Your gnuplot diagrams look better than mine. I'd like to know how you made them.  :-) It's helpful to put some info about how you made things on the image description page so that others can reproduce it, translate labels into their own language, etc. You're also supposed to upload diagrams in SVG format whenever possible, though gnuplot's SVG output doesn't look so great, in my opinion. Example: Image:Waveforms.svg vs Image:Waveforms.png. I'm still uploading them in PNG until either 1. the SVG rendering gets better or 2. I learn to produce better SVGs. — Omegatron 16:20, 13 March 2006 (UTC)

Sample encoding

Doesn't the OGG encoding of the samples use psychoacoustic modelling? And wouldn't that severely distort the result, considering that grey noise is based on psychoacoustics itself? Using uncompressed or losslessly compressed sample would IMO be vastly preferable. --Brazzy 11:13, 30 November 2005 (UTC)

Yep. I have the original wav files still, but Wikipedia only accepts ogg. It's not a bad approximation for now. — Omegatron 14:09, 30 November 2005 (UTC)

forgive my stupidy on the subject.. but in using psychoacoustic modelling surely it would _sound_ the same to a human and that's what's important right? if you were going to use them for anything wouldn't you have software / equipment to generate them? --Streaky 01:34, 11 September 2006 (UTC)

Theoretically, yeah, but psychoacoustic modelling isn't really meant for pure noise, and throws away information that distorts the sound. Apparently you can pack pure WAV files into the ogg container, so I might do that if I get around to it someday and upload it separately... — Omegatron 04:47, 11 September 2006 (UTC)

Layout?

Now that we've got frequency graphs, it would be useful to give more significance to the layout of the page, to ensure that the graphs and their descriptions stay aligned. I'm not a Wiki or HTML expert...any takers on this? --Ktims 02:15, 14 March 2006 (UTC)

You can add {{clear}} after each definition if you want. Is it possible to make the brown and purple fill up the entire x axis? Or do the others look ugly on the same scale? — Omegatron 03:35, 14 March 2006 (UTC)

I just tried it out, and unfortunately increasing the vertical range flattens the pink and brown curves out so much that they look too close. I may recreate the charts with gnuplot's 'smooth bezier' option when I get time, they look a lot less erratic, and for these purposes that's better IMO.

Branchlist

Why do you keep removing it? Its a legit WP construct is it not? Pls resond on Project Electronics as we have the same problem of organising our articles. If we dont start it soon, it will become a massive task--Light current 05:18, 5 April 2006 (UTC)

Shot noise

So where does shot noise fit in this scheme? I have reservations about the color-to-name scheme because I personally have heard only about white noise and pink noise (especially the uniform spectrum part). And what about Johnson noise, for that matter. What if these articles don't have a cited noise-to-color correspondence? --Ancheta Wis 20:41, 18 June 2006 (UTC)

According to Johnson-Nyquist_noise#Explanation, it's white. Shot noise doesn't specify, but this external link says it's white as well. The colors have nothing to do with the process that produced the noise; they're just descriptions of the frequency spectrum produced. — Omegatron 00:42, 19 June 2006 (UTC)

See Talk:Electronic noise. — Omegatron 04:47, 11 September 2006 (UTC)

Origin

How did this whole spectrum → color thing start, anyway? Was "white noise" coined first, to make an analogy to white light with all the colors of the spectrum present? Was "Brownian noise" coined first, named for Brownian motion, but then corrupted into a color abbreviation and followed by others? — Omegatron 17:57, 24 October 2006 (UTC)

American Heritage Dictionary just says: "[From the analogy with white light.]"

From the article:

The color names for these different types of sounds are derived from a loose analogy between the spectrum of frequencies of sound wave present in the sound (as shown in the blue diagrams) and the equivalent spectrum of light wave frequencies. That is, if the sound wave pattern of "blue noise" were translated into light waves, the resulting light would be blue, and so on.

Can we get a reference on that? Brown noise, for instance, was not named for the color, and the colors listed do not actually correspond with the actual color spectrum. — Omegatron 04:37, 4 December 2006 (UTC)

what does 'DC' mean in the explanation of Brown noise?

The term 'DC' is not defined and I could not quickly determine its meaning via google or the disambiguation page here on wikipedia... could any help? Kinser 03:20, 18 January 2007 (UTC)

Direct current, or a signal with a frequency of zero. — Omegatron 21:48, 18 March 2007 (UTC)

Removed Mains hum section

Sorry for being a party pooper, but I've removed the mains hum section. I'm sad to say, but it is not a colour, as the title of the article tells me, and there is an article named Mains hum. Also, it isn't a pure noise (i.e., it isn't "a random signal", as the first paragraph of the article says), I hope everyone is okay with that. I'll move the relevant discussion to Talk:Mains hum now. +mwtoews 08:50, 8 February 2007 (UTC)

Talk content moved. +mwtoews 08:58, 8 February 2007 (UTC)

I agree with that change. — Omegatron 23:31, 18 February 2007 (UTC)

Orange noise and first graders

Funny as some people may have found it, I have removed the comment about orange noise being generated by a class of first graders on plastic recorders. It was a dumb joke, insulting to a real musical instrument, and factually incorrect. Thee strikes and that comment is out. —The preceding unsigned comment was added by Eijkhout (talk • contribs) 13:51, 18 February 2007 (UTC).

It was in the original newsgroup posting. I wonder if the definition itself should be removed. — Omegatron (talk) 15:54, 12 April 2008 (UTC)

structuredness

i thought colors of noise were used to differentiate between levels of structured or erratic noise. did i believe another ghost then?· Lygophile has spoken 17:18, 19 March 2007 (UTC)

Graphs meaning for laymen

Sorry if I'm being dumb here, but I'm struggling to understand how the graphs, particularly of white v pink noise, relate to their descriptions. White noise is described as having a flat frequency spectrum in linear space, but the graph shows a (roughly) flat spectrum with a logarithmic scale. Pink noise is described as being "flat in logarithmic space", but its graph shows a downward trend on the same log scale. I'm guessing that the y-axis of the graphs (dB) is not what they are "flat" in (ie power?), but it all seems horribly confusing. Also the slight rise in the white noise graph is attributed to it being a log rather than linear scale, but that can't be right, if it were on a linear scale it would surely slope downwards dramatically -- so it seems more like the slight rise is an aberration and this isn't true white noise (hardly surprising if it was generated from an OGG file...) Can anyone shed some light? Thanks. 217.169.15.38 19:23, 11 October 2007 (UTC)

Yes, you are quite correct. The plots are obviously wrong, and should be corrected or removed. Furthermore, a grey noise curve can only be correct for one absolute level in dBSPL (e.g. dB(A) is only valid for 40dB@1kHz, or -54dBPa@1kHz) and will look more like the white noise as the absolute level increases. Therefore it is questionable if such a graph is of any help without a more stringent (correct) explanation.— DrD 00:08, 10 November 2007 (UTC)

The y axis of the graphs is logarithmic. dB are logarithmic. Maybe it would be better to have two axes, one log (1, 10, 100, 1000) and the opposing axis in dB, to make this clear.

The x axis is logarithmic, but the measurement is from an FFT, so the measurement bin divisions are linear. I see how this could be confusing, but the angled straight line spectrums will not show up unless they are plotted on a log-log plot.

Hmm... The slight rise in the white noise appears in Adobe Audition, too, which I think was performed on the original wave file. Could be an artifact of Audition's white noise generator (aliasing?), an artifact of the measurement method, or an artifact of the way it's plotted.

Also, is the grey noise completely wrong? I used white noise and filtered it with an inverted A-weighting curve. But maybe it should have started with pink noise?

It would be better if both the signals and the analysis were generated purely in mathematics software (GNU Octave), rather than audio software, but I never got around to this. — Omegatron (talk) 15:38, 12 April 2008 (UTC)

Relevance of analogies between sound and light

In some aspects sound and light are not even considered to have any relationship, but in some cases the two phenomena can be represented with similar models. An example of the latter case is that a sampling in the time domain of the respective field variables of the both phenomena could be Fourier transformed to the frequency domain and be represented as (e.g.) energy distribution vs frequency. Since different distributions of light energy vs frequency can be interpreted as different colours (which is a very obvious interpretation), it is very tempting to do an analogy with sound. But, since sound and light are perceived totally different, it does not make much sense to do a translation of colours to sounds, apart from in some (actually very few) cases:

Evenly distributed noise (in a linear sense, i.e. perfectly random noise) can be called white and the logarithmically distributed counterpart can be called pink in order to suggest that the energy in the low frequency region is dominating in the latter case. The same line of reasoning can be applied for blue light/noise (a domination of high frequency contents) and for black light/noise (absense of energy throughout the whole frequency range, which does not seem as much of a "help" for visualising the nature of the noise - it is kind of obvious anyway). These definitions can be seen as "standard" definitions, are more or less well-established and give a good picture of how the noise is distributed in each case.

But when it comes to the other colours than these mentioned it is not possible to make an obvious and unique interpretation. Neither will these interpretation be of any help. In the best of scenarios we have some academic interpretations that will not help by giving any "feeling" of what kind of noise it is. A nonsense name would be equally sufficient. In the worst scenario we have some childish and far-fetched attempts to make analogies that are both misleading and non-intuitive.

For example: Brown noise is not even an analogy to the corresponding colour. Furthermore, if brown noise is considered as a result of Brownian motion, then the energy distribution is a result of how you implement the random walk amplitude and frequency wise. The resulting distribution can be totally different from what is stated here (the definition here is as meaningful as claiming that the top speed of a car is twice the top speed of a bus). Also the usefulness of such noise can be doubted. Nevertheless, this definition has some merits, mostly because it has been around for some time. Red noise actually implies absence of high frequency contents (dark pink had been more appropriate...) Purple noise is maybe relevant, but it should in that case replace the blue noise definition, which in analogy to the pink noise actually should be called light blue... The grey noise definition contradicts the notion of white noise being perceived as white. Furthermore, its energy distribution should, according to how it is defined here, be a function of the absolute sound pressure level. It will therefore be meaningless nonsense in a file, as a electrical signal and so on unless it is played back at a corresponding calibrated level. These colours, plus orange, green and so on, are all more or less unnecessary constrained constructions.

Thus, if there is no motive to invent new non-standard noise-colour analogies: Please refrain from creating or promoting personal home-made definitions! It only gives too much importance to an analogy which is not inuitive - does anyone really think there is a common notion of, for example, how "green" would sound? –– DrD 22:29, 9 November 2007 (UTC)

These abrownre not "home-made definitions". They are taken right out of Federal Standard 1037C. I changed the "brown" heading to "red" because it fits in more with the colors theme. — Omegatron (talk) 15:57, 12 April 2008 (UTC)

: ??? I had to check in Federal Standard 1037C (http://www.its.bldrdoc.gov/fs-1037/fs-1037c.htm), but I did only find the ordinary "white" and "pink" definitions in their A-Z list. No "brown", "green", "red", "purple" et cetera. Being in the Telecom biz myself, I find it hard to imagine why Federal Standard 1037C should bother to distinguish between e.g. "green", "blue" and "purple" noise. Still I find reason to discuss and write about e.g. "brown" and "red" noise. But my point was that I think there is no merit in taking the sound and light analogy to far.

White noise amplitudes

I started the process of re-generating all the noise audio samples from scratch in GNU Octave. So far I've just got Image:White noise.ogg (uniformly distributed) and Image:Gaussian white noise.ogg (normally distributed).

The Gaussian file sounds about 4.5 dB quieter than the uniformly-distributed file, while SoX says they are actually 8.4 dB different. Is this psychoacoustics or did I measure something wrong? — Omegatron (talk) 04:58, 21 April 2008 (UTC)

Bad reference

^ a b c d e Joseph S. Wisniewski (07 Oct 1996). "Colors of noise pseudo FAQ, version 1.3". comp.dsp. (Web link). Retrieved on 2008-05-04.

The above reference links to an email address...Dbutler1986 (talk) 00:21, 16 July 2008 (UTC)

Huhh?

It would be nice if the article explained what this is all about. Sincerely, GeorgeLouis (talk) 07:08, 24 July 2008 (UTC)

Yes! The article explains the WHAT but it does not explain the WHY. Why are these colors important? How are they used in the audio industry or academia? —Noah 23:57, 13 October 2008 (UTC)

Equal energy versus equal perceptual ammount

White audio noise sounds like a high-frequency hiss; white image noise looks like static and can perceptually be removed with a low-pass filter. Could someone explain the difference between the definition of white noise and this intuitive sense that white noise is a high-frequency noise? I assume it has to do with the difference between energy and amplitude of a frequency band and that blue or violet noise would be perceptually uniform across the spectrum (and so wouldn't be removed easily by a low-pass filter). —Ben FrantzDale (talk) 15:18, 15 September 2008 (UTC)

Too specific

"red noise" has ended up linked from a cliamte page. But the defn here is far far too specific. "red noise" (no-one uses brown) from the climate point of view is just noise with more power at lower frequencies. None of this 6-db-per-whatever stuff William M. Connolley (talk) 21:02, 16 September 2008 (UTC)

Black noise -- a joke?

I'm not questioning whether the entirety of the "black noise" description is a joke, just a part of it. The article says:

Used in modeling various environmental processes. Is said to be a characteristic of "natural and unnatural catastrophes like floods, droughts, bear markets, and various outrageous outages, such as those of electrical power." Further, "because of their black spectra, such disasters often come in clusters."

and cites Manfred Schroeder's book _Fractals, chaos, power laws_. I don't have this book and Amazon won't let me search it. Basically I'm wondering if this description is some kind of joke about power outages during natural disasters.

-- Theclapp (talk) 22:09, 6 January 2009 (UTC)

Here's the quote quoted in another book: Google book search. It even includes the bizarre word choice of 'outrageous'. Binksternet (talk) 22:24, 6 January 2009 (UTC)

Thanks! Theclapp (talk) 13:53, 7 January 2009 (UTC)

Infinite sums can approach a limit

Seeking comments before I strike this sentence:

"An infinite-bandwidth white noise signal is purely a theoretical construct. By having power at all frequencies, the total power of such a signal would be infinite"

Electricmic (talk) 09:05, 3 November 2009 (UTC)