FFT & filtering - newbie needs help

32 views
Skip to first unread message

Huw Roberts

unread,
Sep 19, 1996, 3:00:00 AM9/19/96
to

I am writing a program to analyse WAV file recordings of the human
voice and want to incorporate FFT analysis. I intend to write a
dll in Visual C++ to handle this function and I’m interested in
the range 0 - 2 kHz. I would like a resolution of 0.01 Hz in the
range 0-32 Hz although the resolution in the octaves above can
halve for each higher octave which will mean a resolution of 0.64
Hz in the highest octave.
I am new to FFTs and DSP and not sure of the best way to proceed.
There are 2 methods that I have thought of and would be grateful
for any advice on the practicality of either and general pointers
to good books and web sites to help me get started. In both cases
I will use a sampling rate of 5 to 5.5 kHz depending on what my
SB32AWE can utilise.

Method 1. Use a 512k point FFT to give the 0.01 resolution.
For the higher octaves I average over a number of points to
reduce the resolution. For the highest octave I will be averaging
over 64 points.
Method 2. Use a 8k point FFT to give an initial 0.64Hz
resolution. Halve the sampling rate after first filtering and
then run the 8k FFT to give a 0.32Hz resolution. Repeat this
process until you are reduced to a ~ 80 Hz sampling rate which
gives the 0.01 resolution for the lower range. In all but the
last stage you only keep the data points for the highest octave.
Regards Huw Roberts

pewi

unread,
Sep 20, 1996, 3:00:00 AM9/20/96
to

In article <51sgov$iu9$1...@mhadg.production.compuserve.com>,
1015...@CompuServe.COM says...

>
>I am writing a program to analyse WAV file recordings of the human
>voice and want to incorporate FFT analysis. I intend to write a
>dll in Visual C++ to handle this function and I’m interested in
>the range 0 - 2 kHz. I would like a resolution of 0.01 Hz in the
>range 0-32 Hz although the resolution in the octaves above can
>halve for each higher octave which will mean a resolution of 0.64
>Hz in the highest octave.

Sounds like a wavelet transform is what you need ! One of the main
properties of a wavelet transform is, that you can scale frequency vs time
resolution.

Ciao,
Peter

pe...@microlab.de

Keep the mailer happy


Joseph S. Wisniewski

unread,
Sep 24, 1996, 3:00:00 AM9/24/96
to Huw Roberts

Huw Roberts wrote:

Huw,
I added comp.speech to your newsgroups. It's really the place for this
one.

> I am writing a program to analyse WAV file recordings of the human
> voice and want to incorporate FFT analysis. I intend to write a
> dll in Visual C++ to handle this function and I’m interested in
> the range 0 - 2 kHz. I would like a resolution of 0.01 Hz in the
> range 0-32 Hz

O.K. I'll bite. Why do you want a 0.01Hz resolution for a human voice
analysis? You are talking at least 100 seconds of speech for this
resolution with an FFT, possibly twice that (depending on your window
function). The vocal track can't hold anything even remotly resembling a
steady state (within 0.01Hz) for that length of time. Not even Domingo
could do it! All you are going to get is a big smear.

If you are trying to find instantaneous Formant values or fundamental
pitch with this kind of resolution, the FFT is not the algorithm you
need. You might try LPC analysis, although there are some time-domain
methods that could just extract the pitch, if that's what you are after.

> although the resolution in the octaves above can
> halve for each higher octave which will mean a resolution of 0.64
> Hz in the highest octave.

> I am new to FFTs and DSP and not sure of the best way to proceed.
> There are 2 methods that I have thought of and would be grateful
> for any advice on the practicality of either and general pointers
> to good books and web sites to help me get started. In both cases
> I will use a sampling rate of 5 to 5.5 kHz depending on what my
> SB32AWE can utilise.

8k, 5.5k, and 4k, unless they've "improved" things.

--
Joseph S. Wisniewski | Views expressed are my own, and don't reflect
Ford Motor Company | those of the Ford Motor Co. or affiliates.
Project Sapphire | Trans Am, Daytona, Bonneville, and IROC are
jwis...@ford.com | just races, won by people driving Ford cars!

Johan Halmen UPR

unread,
Sep 26, 1996, 3:00:00 AM9/26/96
to

: > I am writing a program to analyse WAV file recordings of the human

: > voice and want to incorporate FFT analysis. I intend to write a
: > dll in Visual C++ to handle this function and I’m interested in
: > the range 0 - 2 kHz. I would like a resolution of 0.01 Hz in the
: > range 0-32 Hz

Is there anything round 32Hz in a human voice sample? The best I can
do is about 65Hz, in the morning, when my vocal chords are really
relaxed. Domingo won't even do that, I'm a bass and he's a tenor.

Johan Halmen

Joseph S. Wisniewski

unread,
Sep 27, 1996, 3:00:00 AM9/27/96
to Johan Halmen UPR
There's stuff all over the spectrum, the question is, how useful is it?

The white noise components of speech can excite chest cavity resonances
(about 8Hz (if memory server (it hasn't done too well lately))) but the
power levels are going to be very low. Also, under the right conditions,
the vocal chords can do this flap-stick-pause routine that can produce
some suprising low notes.

Off the topic: there are good definitions of white, pink and blue noise,
and I have come across definitions for green, grey, and black noise.
Does anyone know any other colors?

Joseph S. Wisniewski

unread,
Sep 27, 1996, 3:00:00 AM9/27/96
to

Joseph S. Wisniewski wrote:
> Off the topic: there are good definitions of white, pink and
> blue noise, and I have come across definitions for green, grey,
> and black noise.
>
> Does anyone know any other colors?

O.K. I usually don't reply to myself, but email is pouring in asking
what these colors of noise are, so here goes. The first three are
generally accepted terms:

White noise: constant power at any frequency over a band of interest.

Pink noise: power decreases with frequency 3dB/octave, so each octave
contains the same amount of power. Integrated white noise.

Blue noise: power increases with frequency 3dB/octave, or power is
proportional to frequency. Differentiated white noise.

Here's some less conventional ones:

Green noise: supposedly the background noise of the world. A really long
term power spectrum averaged over several outdoor sites. I don't have
the shape of the power spectrum handy. It was defined by some folks
producing relaxation tapes (Mystic Moods, I believe).

Grey noise: noise subjected to a psychoacoustic equal loudness curve
(such as an inverted a-weight curve) so that it sounds like it is
equally loud at all frequencies. I've heard this definition from more
than one person.

I've heard two definitions for the next one:

Black noise 1: whatever comes out of an active noise control system and
cancles an existing noise, leaving the world black. (Jeff Mecure's
definition).

Black noise 2: ultrasonic white noise, with a flat spectrum above 20kHz.
Like a blacklight this is supposed to affect you or your surroundings
without being preceived as sound. (This one came from some sales
literature for an ultrasonic vermin repeller)

Noral D. Stewart

unread,
Sep 27, 1996, 3:00:00 AM9/27/96
to Joseph S. Wisniewski

To my knowledge there has never been an IROC race with a Ford in it.
IROC supplies the same car to all drivers and I do not believe they have
ever used Fords. If someone at Ford has better information, I stand to
be corrected.

Jason Daicheng Chow

unread,
Sep 27, 1996, 3:00:00 AM9/27/96
to

Noral D. Stewart (no...@ix.netcom.com) wrote:
: Joseph S. Wisniewski wrote:
: >

Uhmmmm.... so what... My dog slurps his own vomit and my cat purrs when
I kick it.

--
____________________________________
| Jason Daich Chow |
|====================================|
|http://www.prism.gatech.edu/~gt7060a|
| All generalizations are false, |
| including this one. |
|- - - - - - - - - - - - - - - - - - |
/)| da...@cc.gatech.edu |(\
/ )|____________________________________|( \
__( (|____________________________________|) )__
((( \ \ ) /_) (_\ ( / / )))
(\\\ \ \_/ / \ \_/ / ///)
\ / \ /
\ _/ \_ /
/ / \ \
/ / \ \

Kevin Max Krebs

unread,
Sep 27, 1996, 3:00:00 AM9/27/96
to

Joseph S. Wisniewski wrote:
> Off the topic: there are good definitions of white, pink and
> blue noise, and I have come across definitions for green, grey,
> and black noise.
>
> Does anyone know any other colors?


Apart from the colours you listed in your previous message, I have
found reference to "Brown Noise." The following is quoted from the
help-file of CoolEdit 95 (An incredible share-ware wave editor):

Brown noise has a spectral frequency of 1/f^2. Which means, in
English, that there is much more low-end, low-frequency components to
the noise, which results in thunder and waterfall like sounds. Brown
noise is called that because, when viewed, the wave follows a
Brownian motion curve. That is, the next sample in the waveform is
equal to the previous sample, plus a small random amount. This gives
the appearance of a mountain range when graphed. The wave pattern is
very predictable.

There is, however, no source listed that I can find for this
information.


Gints Klimanis

unread,
Sep 27, 1996, 3:00:00 AM9/27/96
to

As a personal preference, I call 1/f^2 noise red. If pink noise is 1/f,
1/f^2 should be red noise. I read that in an acoustics text a decade
ago, but I can't recall the title. A pure 1/f^2 noise would be the
average of the current sample plus the previous sample, a one zero filter:
y[n] = 0.5*(x[n] + x[n-1); The spectral response of that would be
a down-sloping pattern rather than a mountain range.

John W. Herman

unread,
Sep 28, 1996, 3:00:00 AM9/28/96
to

"Joseph S. Wisniewski" <jwis...@ford.com> writes:

>Off the topic: there are good definitions of white, pink and blue noise,
>and I have come across definitions for green, grey, and black noise.
>Does anyone know any other colors?

Can you explain? I understand blue and green (high and middle parts of the
spectrum) but grey and black escape me.

pro...@connix.com

unread,
Sep 29, 1996, 3:00:00 AM9/29/96
to

In <1996Sep28....@nosc.mil>, her...@nosc.mil (John W. Herman) writes:
>"Joseph S. Wisniewski" <jwis...@ford.com> writes:
>
>>Off the topic: there are good definitions of white, pink and blue noise,
>>and I have come across definitions for green, grey, and black noise.
>>Does anyone know any other colors?
>

Oceanic ambient noise (ie, noise distant from the sources) is often
described as "red" due to the selective absorption of higher frequencies.


P.J. "Josh" Rovero email: pro...@connix.com


Jon M. Risch

unread,
Sep 29, 1996, 3:00:00 AM9/29/96
to Joseph S. Wisniewski

Joseph S. Wisniewski wrote:
>
> Joseph S. Wisniewski wrote:
> > Off the topic: there are good definitions of white, pink and
> > blue noise, and I have come across definitions for green, grey,
> > and black noise.
> >
> > Does anyone know any other colors?
>

I have some more colors of noise. I have heard of brown noise, which is 3
dB/octave bass heavy compared to pink noise.

Strictly speaking, an integration or differentiation results in 6 dB/oct
slopes from the original, not 3 dB/oct slopes.

Therefore, brown noise would be integrated white noise, not pink as you
stated.

Using these directions and the colors already defined, perhaps differentiated
white noise (rising 6 dB/octave with respect to white noise) would be purple
(or violet) noise?

JR

Charles H. Mosher

unread,
Sep 29, 1996, 3:00:00 AM9/29/96
to

Joseph S. Wisniewski (jwis...@ford.com) wrote:
: Joseph S. Wisniewski wrote:
: > Off the topic: there are good definitions of white, pink and
: > blue noise, and I have come across definitions for green, grey,
: > and black noise.
: >

: White noise: constant power at any frequency over a band of interest.

Correct.

: Pink noise: power decreases with frequency 3dB/octave, so each octave


: contains the same amount of power. Integrated white noise.

Integrated white noise would fall off at 6 db per octave. Pink noise can
be derived from white noise, but the shaping network is somewhat more
complicated. General Radio provided a pink noise spectrum shaper with one
of their white noise generators back in the 1950s. This can get
interesting and complicated; if you wish to learn more, please email me.

: Blue noise: power increases with frequency 3dB/octave, or power is


: proportional to frequency. Differentiated white noise.

Differentiated white noise would have power increasing at 6 db per octave.
The shaping/equalizing problem is similar to that of pink noise above.
Blue noise would be differentiated PINK noise.

Don't know about the other colors.

Enjoy.

--
Charles Mosher
ratr...@alumni.caltech.edu for reliability
ratr...@svpal.org is most convenient for me.


Ralf W. Stephan

unread,
Sep 29, 1996, 3:00:00 AM9/29/96
to

Kevin Max Krebs writes:
> [1/f^2 noise]

> There is, however, no source listed that I can find for this
> information.

I have seen a definition of 1/f^2 noise in "The Science Of Fractal
Images", the sequel to "The Beauty Of Fractals", both Springer-Verlag.
Don't have it handy, sorry.


ralf
--
Lynx-enhanced pages at http://www.bayreuth-online.de/~stephan

Joseph S. Wisniewski

unread,
Sep 30, 1996, 3:00:00 AM9/30/96
to

Got lots of feedback on this one, so here's the revised colors of noise
list. Thanks to the many people who pointed out the flaws in my pink and
blue definitions. Thanks Kev fot the pointer to FS-1037C. The noises are
now in spectral order (artistic license has been taken over where white,
black, grey, and brown fit into a spectrum). Anyone is welcome to help
fill in the gaps. Thanks also the the person who sent in the definition
of orange noise, but that pun is too bad to use.

--------------------------------------------------------------------
Colors of noise, version 1.1

White noise (common definition) constant power at any frequency over a
given (finite) range of frequencies.

Pink noise (common definition) power decreases with frequency 3dB/octave
over a given range of frequencies (which does not include DC), so each


octave contains the same amount of power.

Red noise (defined by P.J. "Josh" Rovero) oceanic ambient noise (ie,


noise distant from the sources) is often described as "red" due to the
selective absorption of higher frequencies.

Green noise (defined by some folks producing relaxation tapes, Mystic
Moods, I believe) supposedly the background noise of the world. A really


long term power spectrum averaged over several outdoor sites. I don't
have the shape of the power spectrum handy.

**** I now have two blue noise definitions *****

Blue noise (common definition) power increases with frequency 3dB/octave
over a given (finite) range of frequencies, power is proportional to
frequency. This can be good noise for dithering. No one disputed this
definition, but someone suggested a pointer to (FS-1037C) which has a
different definition.

Blue noise (from FS-1037C) In a spectrum of frequencies, a region in
which the spectral density, i.e. , power per hertz, is proportional to
the frequency. By this definition, blue noise is differentiated white
noise. Jon Risch suggests calling differentiated white noise purple or
violet noise.

Grey noise (heard this one a couple of times, but can't put my finger on
a source) noise subjected to a psychoacoustic equal loudness curve
(such as an inverted a-weight curve) over a given range of frequencies,
so that it sounds like it is equally loud at all frequencies. This would
be a better definiton of "white noise" than the "equal power at all
frequencies" definition, since real "white light" has the power spectrum
of a 5400K black body, not an equal power spectrum. I've heard this


definition from more than one person.

Brown noise (Jon M. Risch) power decreases with frequency 6dB/octave
over a given range of frequencies (which does not include DC). This is
the result of integrating white noise.

**** Two different definitions of black (silent) noise ****

Black noise (Jeff Mercure) whatever comes out of an active noise control
system and cancles an existing noise, leaving the world world noise
free.

Black noise (seen in the sales literature for an ultrasonic vermin
repeller) ultrasonic white noise, with a flat spectrum above 20kHz.
This black noise is like the so-called "black light" with frequencies
too high to be preceived as sound, but still capable of affecting you or
your surroundings. (The comic book character "Iron Man" used to have a
"black light beam" that could darken a room like this ;-)

1/f^2 noise is mentioned by Kevin Max Krebs, but without a color!

Anyone got a good authorative definition for blue noise?

dau...@gwdvms.gwdg.de

unread,
Oct 1, 1996, 3:00:00 AM10/1/96
to

>>>Off the topic: there are good definitions of white, pink and blue noise,
>>>and I have come across definitions for green, grey, and black noise.
>>>Does anyone know any other colors?
>>
>
I (and I think others) wouldlove to see the complete list when you have
collected them all. Thanks in advance.
David Auerbach

Greg J Szekeres

unread,
Oct 1, 1996, 3:00:00 AM10/1/96
to


Yes, I would like to see a colorfull FAQ.

greg


Joseph S. Wisniewski

unread,
Oct 2, 1996, 3:00:00 AM10/2/96
to Greg J Szekeres

Colors of noise pseudo FAQ, version 1.2
------------------------------------------------------------------------
That email just keeps coming in. So, here's the latest rev. Thanks to

the many people who pointed out the flaws in my pink and blue
definitions. Thanks Kev fot the pointer to FS-1037C. Due to popular
demand, I am reversing my previous stand and adding the definition of
orange noise. You will be sorry.

The noises are now in spectral order (artistic license has been taken
over where white, black, grey, and brown fit into a spectrum). Anyone is

welcome to help fill in the gaps, especially a stronger definition of
blue.
------------------------------------------------------------------------

White noise (common definition) power density is constant over a finite
frequency range. AKA Johnson noise.

Pink noise (common definition) power density decreases 3dB per octave
with increasing frequency (density proportional to 1/f) over a finite
frequency range which does not include DC. Each octave contains the same
amount of power. Many point out that this is not a trivial filtering
problem. AKA flicker noise.

Red noise (defined by P.J. "Josh" Rovero) oceanic ambient noise (ie,
noise distant from the sources) is often described as "red" due to the

selective absorption of higher frequencies. (Anyone have the spectrum?)

Orange noise (anonymous definition) quasi-stationary noise with a
finite power spectrum with a finite number of small bands of zero
energy dispersed throughout a continuous spectrum. These bands of
zero energy are centered about the frequencies of musical notes in
whatever system of music is of interest. Since all in-tune musical
notes are eliminated, the remaining spectrum could be said to consist
of sour, citrus, or "orange" notes. Orange noise is most easily
generated by a roomfull of primary school students equipped with
plastic soprano recorders. (Anyone foolish enough to want the
spectrum?)

Green noise (defined by some folks producing relaxation tapes, Mystic
Moods, I believe) supposedly the background noise of the world. A really

long term power spectrum averaged over several outdoor sites. Rather
like pink noise with a hump added around 500Hz. (Anyone have the
spectrum?)

Blue noise (FS-1037C) power density increases 3dB per octave with
increasing frequency (density proportional to f) over a finite frequency
range. This can be good noise for dithering.

Purple noise (suggestion from Jon Risch) power density increases 6dB
per octave with increasing frequency (density proportional to f^2) over
a finite frequency range. Differentiated white noise. AKA violet noise.

Grey noise (heard this one a couple of times, but can't put my finger on
a source) noise subjected to a psychoacoustic equal loudness curve
(such as an inverted a-weight curve) over a given range of frequencies,
so that it sounds like it is equally loud at all frequencies. This would
be a better definiton of "white noise" than the "equal power at all
frequencies" definition, since real "white light" has the power spectrum
of a 5400K black body, not an equal power spectrum.

Brown noise (Jon M. Risch, rbmccammon) power density decreases 6dB per
octave with increasing frequency (density proportional to 1/f^2) over a
frequency range which does not include DC. Is not named for a power
spectrum that suggests the color brown, rather, the name is a coruption
of Brownian motion. If we were going to pick a color, red might be good
since pink noise lies between this noise and white noise. Unfortuantly,
red is already taken. AKA "random walk" or "drunkard's walk" noise.

**** Two different definitions of black (silent) noise ****

Black noise (Jeff Mercure) whatever comes out of an active noise control
system and cancles an existing noise, leaving the world world noise

free. (The comic book character "Iron Man" used to have a "black light
beam" that could darken a room like this, and popular SCI-FI has an
annoying tendancy to portray active noise control in this light.)

Black noise (seen in the sales literature for an ultrasonic vermin

repeller) power density is constant for a finite frequency range above
20kHz. Ultrasonic white noise. This black noise is like the so-called


"black light" with frequencies too high to be preceived as sound, but
still capable of affecting you or your surroundings.

M. Arnao

unread,
Oct 2, 1996, 3:00:00 AM10/2/96
to

>**** Two different definitions of black (silent) noise ****
>
>Black noise (Jeff Mercure) whatever comes out of an active noise control
>system and cancles an existing noise, leaving the world world noise
>free. (The comic book character "Iron Man" used to have a "black light
>beam" that could darken a room like this, and popular SCI-FI has an
>annoying tendancy to portray active noise control in this light.)
>
>Black noise (seen in the sales literature for an ultrasonic vermin
>repeller) power density is constant for a finite frequency range above
>20kHz. Ultrasonic white noise. This black noise is like the so-called
>"black light" with frequencies too high to be preceived as sound, but
>still capable of affecting you or your surroundings.
>
>
>--
> Joseph S. Wisniewski | Views expressed are my own, and don't reflect
> Ford Motor Company | those of the Ford Motor Co. or affiliates.
> Project Sapphire | Trans Am, Daytona, Bonneville, and IROC are
> jwis...@ford.com | just races, won by people driving Ford cars!

according to manfred schroeder, in "fractals, chaos, power laws," black noise
has an f ^ -beta spectrum, with beta > 2, and is 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."

Clay Shumate Turner

unread,
Oct 3, 1996, 3:00:00 AM10/3/96
to

Another definition of black noise pertains to a dithering method used
by Telecom Australia with paging 1200 bps POCSAG. Basically in a
simulcast radio system, one worries about the overlap areas where the
interference creates nulls. The standard approach is to offset each
transmitter's frequency by different albeit small amounts. This ensures
that the nulls are moving. But this requires careful frequency offset
management.
Telecom's answer was to add a time varying
pseudo random frequency offset. Each transmitter has its own pseudo
random generator with unique period and/or starting time. This resulted
in the nulls having a chaotic motion which works quite well in this
application. Another advantage of this method is the lessoning of the
frequency stability requirements of the oscillators in the transmitters.

Clay S. Turner

Frederick Wahl

unread,
Oct 4, 1996, 3:00:00 AM10/4/96
to

Joseph S. Wisniewski (jwis...@ford.com) wrote:
: Got lots of feedback on this one, so here's the revised colors of noise
: list. Thanks to the many people who pointed out the flaws in my pink and
: blue definitions. Thanks Kev fot the pointer to FS-1037C. The noises are

: now in spectral order (artistic license has been taken over where white,
: black, grey, and brown fit into a spectrum). Anyone is welcome to help
: fill in the gaps. Thanks also the the person who sent in the definition

: of orange noise, but that pun is too bad to use.

Gee, it sure would be great is someone would upload a few examples of
these in the form of short .WAV files (short samples cold be looped).
Pink noise especially.

: --------------------------------------------------------------------


: Colors of noise, version 1.1

: White noise (common definition) constant power at any frequency over a
: given (finite) range of frequencies.

: Pink noise (common definition) power decreases with frequency 3dB/octave
: over a given range of frequencies (which does not include DC), so each
: octave contains the same amount of power.

: Red noise (defined by P.J. "Josh" Rovero) oceanic ambient noise (ie,


: noise distant from the sources) is often described as "red" due to the
: selective absorption of higher frequencies.

: Green noise (defined by some folks producing relaxation tapes, Mystic


: Moods, I believe) supposedly the background noise of the world. A really

: long term power spectrum averaged over several outdoor sites. I don't


: have the shape of the power spectrum handy.

: **** I now have two blue noise definitions *****

: Blue noise (common definition) power increases with frequency 3dB/octave
: over a given (finite) range of frequencies, power is proportional to

: frequency. This can be good noise for dithering. No one disputed this


: definition, but someone suggested a pointer to (FS-1037C) which has a
: different definition.

: Blue noise (from FS-1037C) In a spectrum of frequencies, a region in
: which the spectral density, i.e. , power per hertz, is proportional to
: the frequency. By this definition, blue noise is differentiated white
: noise. Jon Risch suggests calling differentiated white noise purple or

: violet noise.

: Grey noise (heard this one a couple of times, but can't put my finger on
: a source) noise subjected to a psychoacoustic equal loudness curve
: (such as an inverted a-weight curve) over a given range of frequencies,
: so that it sounds like it is equally loud at all frequencies. This would
: be a better definiton of "white noise" than the "equal power at all
: frequencies" definition, since real "white light" has the power spectrum

: of a 5400K black body, not an equal power spectrum. I've heard this


: definition from more than one person.

: Brown noise (Jon M. Risch) power decreases with frequency 6dB/octave

: over a given range of frequencies (which does not include DC). This is


: the result of integrating white noise.

: **** Two different definitions of black (silent) noise ****

: Black noise (Jeff Mercure) whatever comes out of an active noise control
: system and cancles an existing noise, leaving the world world noise
: free.

: Black noise (seen in the sales literature for an ultrasonic vermin
: repeller) ultrasonic white noise, with a flat spectrum above 20kHz.
: This black noise is like the so-called "black light" with frequencies


: too high to be preceived as sound, but still capable of affecting you or

: your surroundings. (The comic book character "Iron Man" used to have a
: "black light beam" that could darken a room like this ;-)

: 1/f^2 noise is mentioned by Kevin Max Krebs, but without a color!

: Anyone got a good authorative definition for blue noise?

: --

: Joseph S. Wisniewski | Views expressed are my own, and don't reflect
: Ford Motor Company | those of the Ford Motor Co. or affiliates.
: Project Sapphire | Trans Am, Daytona, Bonneville, and IROC are
: jwis...@ford.com | just races, won by people driving Ford cars!

Thanks in advance.
Rick

Roy McCammon

unread,
Oct 5, 1996, 3:00:00 AM10/5/96
to

Frederick Wahl wrote:
>
> Joseph S. Wisniewski (jwis...@ford.com) wrote:

> : Blue noise (common definition) power increases with frequency 3dB/octave
> : over a given (finite) range of frequencies, power is proportional to
> : frequency. This can be good noise for dithering. No one disputed this
> : definition, but someone suggested a pointer to (FS-1037C) which has a
> : different definition.

> : Blue noise (from FS-1037C) In a spectrum of frequencies, a region in
> : which the spectral density, i.e. , power per hertz, is proportional to
> : the frequency. By this definition, blue noise is differentiated white
> : noise.

These are the same. 3dB/octave means: when the frequency dpoubles,
the power doubles, i.e. the power is proportional to frequncy.

Neither of these are differentiated white noise, which would give 6 dB/Octave.
Differentiating white noise would give an rms value that was proportional to
frequency and a power that was proportional to the square of frequency.

Opinions expressed herein are my own and may not represent those of my employer.

RettaMitnA

unread,
Oct 5, 1996, 3:00:00 AM10/5/96
to

One of the best (possibly the absolute best) sound editors for
Windows (versions for 3.1 and 95) can generate Pink, White, and Brown
noise. Try downloading Cool Edit from http://www.netzone.com/syntrillium.

-retta...@aol.com (Joel Hardy)

Joseph S. Wisniewski

unread,
Oct 7, 1996, 3:00:00 AM10/7/96
to

Colors of noise pseudo FAQ, version 1.3
------------------------------------------------------------------------
That email just keeps coming in. So, here's the latest rev. Thanks to

the many people who pointed out the flaws in my pink and blue
definitions. Thanks Kev fot the pointer to FS-1037C. Due to popular
demand, I am reversing my previous stand and adding the definition of
orange noise.

The noises are now in spectral order (artistic license has been taken


over where white, black, grey, and brown fit into a spectrum). Anyone is

welcome to help fill in the gaps. We're up to three defintions of black
noise. Keep them coming!
------------------------------------------------------------------------

White noise (common definition) power density is constant over a finite
frequency range. AKA Johnson noise.

Pink noise (common definition) power density decreases 3dB per octave
with increasing frequency (density proportional to 1/f) over a finite

frequency range which does not include DC. Each octave contains the
same amount of power. Many point out that this is not a trivial


filtering
problem. AKA flicker noise.

Red noise (common definition within the oceanographic field, contributed
by P.J. "Josh" Rovero) (Anyone have the spectrum?)

oceanic ambient noise (ie, noise distant from the sources) is
often described as "red" due to the selective absorption of
higher frequencies."

Orange noise (anonymous contribution) (Anyone foolish enough to want
the spectrum?)

quasi-stationary noise with a finite power spectrum with a
finite number of small bands of zero energy dispersed throughout
a continuous spectrum. These bands of zero energy are centered
about the frequencies of musical notes in whatever system of music
is of interest. Since all in-tune musical notes are eliminated,
the remaining spectrum could be said to consist of sour, citrus,
or "orange" notes. Orange noise is most easily generated by a
roomfull of primary school students equipped with plastic soprano
recorders.

Green noise (defined by some folks producing relaxation tapes, Mystic


Moods, I believe) supposedly the background noise of the world. A really

long term power spectrum averaged over several outdoor sites. Rather
like pink noise with a hump added around 500Hz. (Anyone have the
spectrum?)

Blue noise (FS-1037C) power density increases 3dB per octave with
increasing frequency (density proportional to f) over a finite frequency

range. This can be good noise for dithering.

Purple noise (origional definition, contributed by Jon Risch) power


density increases 6dB per octave with increasing frequency (density
proportional to f^2) over a finite frequency range. Differentiated

white noise. AKA violet noise.

Grey noise (heard this one a couple of times, but can't put my finger on
a source) noise subjected to a psychoacoustic equal loudness curve
(such as an inverted a-weight curve) over a given range of frequencies,
so that it sounds like it is equally loud at all frequencies. This would
be a better definiton of "white noise" than the "equal power at all
frequencies" definition, since real "white light" has the power spectrum
of a 5400K black body, not an equal power spectrum.

Brown noise (Jon M. Risch, rbmccammon) power density decreases 6dB per


octave with increasing frequency (density proportional to 1/f^2) over a
frequency range which does not include DC. Is not named for a power
spectrum that suggests the color brown, rather, the name is a coruption
of Brownian motion. If we were going to pick a color, red might be good
since pink noise lies between this noise and white noise. Unfortuantly,
red is already taken. AKA "random walk" or "drunkard's walk" noise.

**** Three different definitions of black (silent) noise ****

Black noise (contributed by Jeff Mercure, his own definition) whatever


comes out of an active noise control system and cancles an existing
noise,

leaving the world world noise free. (The comic book character "Iron
Man"
used to have a "black light beam" that could darken a room like this,


and popular SCI-FI has an annoying tendancy to portray active noise
control in this light.)

Black noise (seen in the sales literature for an ultrasonic vermin


repeller) power density is constant for a finite frequency range above

20kHz. Ultrasonic white noise. This black noise is like the so-called


"black light" with frequencies too high to be preceived as sound, but
still capable of affecting you or your surroundings.

Black noise (Manfred Schroeder, "fractals, chaos, power laws,"
contributed
by Mike Arnao)

has an f ^ -beta spectrum, with beta > 2, and is 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."

--

Andrew

unread,
Oct 8, 1996, 3:00:00 AM10/8/96
to

In article <Pine.GSO.3.95.961003...@panther.Gsu.EDU>
phy...@panther.Gsu.EDU
Clay Shumate Turner writes:


The problems of nulls and multipaths were seen as an obstacle to
lower power digital radio transmissions. One approach which is
successful in elimination of the problem is to use a large number
of tightly packed carriers instead of the single carrier normally used.

The technique is known as Coded Orthogonal Frequency Division
Multiplex (COFDM). The result has been demonstrated to be so good
that digital hi-fi quality broadcasts can be received easily in city
cars without sound degradation. I believe that the receivers are still
rather large at present.

Andrew Silverman
--
Env...@measure.demon.co.uk

Patrick Piecha

unread,
Oct 8, 1996, 3:00:00 AM10/8/96
to

jwis...@ford.com (Joseph S. Wisniewski) wrote:

> Colors of noise pseudo FAQ, version 1.3

Very good FAQ.
It would be pleasant to see pseudocode-algorithms to create all these
kinds of noise as raw audio data for measuring purposes. How about?


Regards, Patrick

--
P.Pi...@Airport.in-berlin.de (preferred)
s015...@rz.fhtw-berlin.de
PGP available

p...@globalnet.co.uk

unread,
Oct 13, 1996, 3:00:00 AM10/13/96
to

"Joseph S. Wisniewski" <jwis...@ford.com> wrote:

>Colors of noise pseudo FAQ, version 1.3
>------------------------------------------------------------------------
>That email just keeps coming in. So, here's the latest rev. Thanks to
>the many people who pointed out the flaws in my pink and blue
>definitions. Thanks Kev fot the pointer to FS-1037C. Due to popular
>demand, I am reversing my previous stand and adding the definition of
>orange noise.

>The noises are now in spectral order (artistic license has been taken
>over where white, black, grey, and brown fit into a spectrum). Anyone is
>welcome to help fill in the gaps. We're up to three defintions of black
>noise. Keep them coming!
>------------------------------------------------------------------------

I would be interested to know of any sites holding .wav files
where samples of each colour could be downloaded for listening
using a p.c. soundcard. Many thanks.
--
P.A.Newman <p...@globalnet.co.uk


Reply all
Reply to author
Forward
0 new messages