Convert white noise to brown

"Bill Shilelds"

It is produced from white noise by a - 6dB per octave filter.

..... Phil

Convert white noise to brown what?

VLV

Phil is correct; ignore the other comments. White noise is equal-energy-per-octave, and never spans all frequencies; if you add an upper frequency limit, that 'white' noise becomes 'pink'.

The -6dB per octave filter gives red or brown noise, but there's limits... it's only effective in a limited range of frequencies. Think on it; if you go down to DC from 1 kHz, that's infinitely many octaves, thus infinite energy. And, of course, from 1 kHz to 1 GHz there's twenty octaves of diminution... just how low-noise IS your circuitry, can you do that effectively?

An integrator (resistor, op amp, capacitor in negative feedback) is a near-ideal -6.0206 dB per octave filter (it's not really '6' dB per octave, that's just the casual approximation).

I would like to see how close I can come using passives.

Can anyone provide the configuration and component values (starting point) for a suitable RC network?

Bill Shields

What bandwidth, gain?

Tim

"whit3rd" "Phil Allison"

Phil is correct; ignore the other comments.

I built an audio band pink noise generator about 25 years ago and can well remember the hassle of getting the -3 dB octave slope from 10Hz to 50 kHz accurate.

The white noise source used was a 8 pin DIL pseudo-random number IC made by National Semi for the purpose - long obsolete now. It tended to make a weird steam train like noise rather than pure white.

For accurate white and pink noise, I have relied on a test CD since 1984.

Denon Audio Technical CD: 38C39 -7147

White noise is equal-energy-per-octave, and never spans all frequencies; if you add an upper frequency limit, that 'white' noise becomes 'pink'.

To convert white noise to pink requires that the whole range be filtered so there is equal energy per octave.

An integrator (resistor, op amp, capacitor in negative feedback) is a near-ideal -6.0206 dB per octave filter (it's not really '6' dB per octave, that's just the casual approximation).

..... Phil

"Bill Shilelds" whit3rd

..... Phil

That gives a corner frequency of 15.9Hz with low source impedance, being sub-audio.

Is that correct??

Bill Shields

"Bill Shilelds" "Phil Allison"

The buffer gives you a low source impedance.

If you use a TL071, then add a 220 ohm resistor at the output to prevent instability with capacitive loads.

The noise will be close to Brown down to about 30 Hz.

.... Phil

On a sunny day (Mon, 15 Mar 2010 17:50:24 -0700) it happened John Larkin wrote in :

Why not? Digitally you can multiply with any thing over a fixed interval.

I would like it noted that a -3dB/octave filter converts white noise to pink noise, not to brown noise. To convert white noise to brown/Brownian/red noise, what is needed is -6dB/octave. That is easier, only requiring an integrator, or a approximation of one using an RC lowpass filter (already mentioned in this thread).

Meanwhile, there are white noise to pink noise filters achieving

-3dB/octave to something like +/- 1/2 or 1/4 dB from 20 to 20,000 Hz or so, easy to find by web searching.

- Don Klipstein ( snipped-for-privacy@misty.com)

Still, not quite ideal; there's no DC negative feedback in the filter as described... and there's some terms, like crossover distortion, that might be an issue at higher frequencies. So, it'll drift near the power rail and teach someone why the LM324 is not recommended for audio...

LM324 crossover distortion can be tamed with an output resistor to ground (or -ve on split rails) to convert the output to Class A.

But I did that for LVDT signal conditioner, not audio for listening.

Grant.

"whitless" "Phil Allison"

Still, not quite ideal;

Such a filter = one with a nominal -6dB / oct slope IS ideal in terms of its slope.

..... Phil