What Exactly is Uncorrelated Pink Noise ?

[Tim Link]

I have been trying to understand the tonal effects of mixing stereo channels together with one inverted. Obviously anything that's exactly the same in both channels will disappear, with only the difference remaining. Recently I tried listening to uncorrelated stereo pink noise, first with both channels mixed into the speaker (L+R), and then with the difference, (L-R). I can switch between the two instantly with a click on the computer. What I hear is absolutely no difference what so ever! I didn't expect that. So how do they generate a signal that sounds the exact same when added to itself or subtracted from itself;
L-R = L+R ?

Somehow they have zero correlation what-so-ever and yet sound the same.
Is it 90 degrees out of phase?

 

[Blumlein 88]

Correlated noise would be the same in both channels. L+R would be twice the amplitude. L-R would be silence. Noise is not correlated with itself. It is random. So unless artificially generated noise for any given time will randomly differ from any other given time. Software can also generate the noise that is uncorrelated. So at random any given moment one noise source will be lower and another higher and it will vary randomly and the result of mixing them is also random so it is still noise. If truly uncorrelated, switching phase of noise doesn't change this. The value of the noise will differ at any given moment, but the signal is still randomly varying over the same range. So still noise.

Does this make any sense or am I explaining it poorly?

 

[DonH56]

What @Blumlein 88 said. Uncorrelated stereo noise means each channel gets its own, unique, noise signal. You could try splitting a single (mono) noise source to both channels, in software or by using a Y splitter to the L and R channels, and should be able to hear the difference.

 

[Tim Link]

Correlated noise would be the same in both channels. L+R would be twice the amplitude. L-R would be silence. Noise is not correlated with itself. It is random. So unless artificially generated noise for any given time will randomly differ from any other given time. Software can also generate the noise that is uncorrelated. So at random any given moment one noise source will be lower and another higher and it will vary randomly and the result of mixing them is also random so it is still noise. If truly uncorrelated, switching phase of noise doesn't change this. The value of the noise will differ at any given moment, but the signal is still randomly varying over the same range. So still noise.

Does this make any sense or am I explaining it poorly?

Yes, that makes sense. I guess I just didn't expect the uncorrelation to be so perfect! I supposed that the same signal of any sound that's been doubled and then one signal shifted 90 degrees would sound the same if subtracted or added together - unless a phase shift was audibly detectable.

 

[Tim Link]

What @Blumlein 88 said. Uncorrelated stereo noise means each channel gets its own, unique, noise signal. You could try splitting a single noise source to both channels, in software or by using a Y splitter to the L and R channels, and should be able to hear the difference.

I can play each channel separately in stereo and hear that it's a stereo signal rather than mono, if that's what you mean. I just thought there would randomly be some stuff missing or the same that would cause an abrupt and noticeable change in the sound when I flipped the phase of one channel. I hear absolutely zilch. Not the slightest hint that anything happened at all. With correlated pink noise it goes from all on to nothing, pure silence, if I flip the polarity on one side. It's a perfect cancel, as expected. With decorrelated I thought there'd at least be a slight difference. But no. It's really, really perfectly decorrelated. A perfect no change at all!

 

[DonH56]

Yes, that makes sense. I guess I just didn't expect the uncorrelation to be so perfect! I supposed that the same signal of any sound that's been doubled and then one signal shifted 90 degrees would sound the same if subtracted or added together - unless a phase shift was audibly detectable.

Maybe if it was pure sine waves, but noise is random from one sample (instant in time) to the next, in amplitude and "phase" (phase is pretty meaningless for a noise signal). Uncorrelated should be just that, unrelated to anything else, and that is exactly what you want. There will undoubtedly be random points in time where they cancel, and add, but those are probably (should be) far too fleeting to hear.