Is the Multi OSCs Noise true white noise?

oddson · 8 Mar 2007

At the risk of getting Stefan going on determinism and randomness... nothing generated by a deterministic process is truely random. There was a thread somewhere here where this question was raised, that is: how good an approximation of randomness is it? The conclusion was 'good enough.'

angstrom · 8 Mar 2007

I think he means - is it flat across the spectrum, IE useful for deriving room analysis (etc) from. Obviously other noises are not flat across the spectrum.

that was my understanding anyway.

PaulRiedel · 8 Mar 2007

angstrom wrote: I think he means - is it flat across the spectrum, IE useful for deriving room analysis (etc) from. Obviously other noises are not flat across the spectrum.

that was my understanding anyway.

Exactly what I mean! - Do you know the answer? 😉

stefan · 8 Mar 2007

True white noise? I'd have to say no. And not for nitpicking reasons Oddson 🙂

For almost all applications it's probably close enough to being white noise, but it's not really even aimed towards being properly "white". It's just noise, generated by sine-phase-feedback.

oddson · 8 Mar 2007

If it appears to generate all frequencies evenly then it is white noise, right?

Showing a signal is not white noise is pretty easy, just point to where it's generating significantly more of one frequency than another. But I don't know how you show something is white noise other than pointing to a reasonably flat frequency distribution. 😕

Is there any reason to believe this method would favour one frequency over another?

PaulRiedel · 10 Mar 2007

Stefan, what do you say about that? I must say that oddsons theory seems quite plausible...

stefan · 11 Mar 2007

Yeah. But if there's a -1dB dip somewhere or whatever, don't hold it against me! 🙂

PaulRiedel · 11 Mar 2007

stefan wrote: Yeah. But if there's a -1dB dip somewhere or whatever, don't hold it against me! 🙂

Haha! I'ma quote you on that one in my thesis 😉

infuzion · 12 Mar 2007

I imagine one can test it out your self by doing A/B tests with a white-noise generator that is considered by Pros to be standard with a spectrogram & one of those EQ-imprint plugins (an EQ that can "clone" the EQ curve to one song & apply it to another).

The only thing I'm not crazy @ SM's noise generator is that the seed isn't random, but that's easily taken care of by hooking up a Random up to the phase.

oddson · 12 Mar 2007

I just tried this

streamin in; streamout out; float mem[1000000]=rand(-1,1); float index; out = mem[index]; index = index + 1; index=index-(index>1000000)&1000000;

With the poly to graph and FFT display it actually looks not as good as the phase overdrive method.

Edit - on second look they seem about the same...