noise not so noisy?

Floating Point's icon

Hey I discovered something that needs explaining.

If you sample the output from noise~ at a rate equal to 2^N samples, you get a regular patternation. Same in gen as well. The patterns alter when you alter the sig vector and/or the IO vector too.

Max Patch
Copy patch and select New From Clipboard in Max.

It seems to me that the noise generating algorithm is not so random. Have a look at the patch below to see what I mean. At first I thought it was just the calcount of scope doing some sort of visual aliasing, but from my little sonification using cycle~ there is evidently more to it than 'meets the eye'. Can anyone account for this?

I haven't tried it using pink~, maybe someone can check that too...

Peter Castine's icon

Indeed.

I've written in the past (ie Max 4 days) about various issues with [random] and [noise~]. The core RNG used then was taken from an algorithm published in Numerical Recipes, which was fairly fast but not a serious random number generator. I'm not sure if this algorithm is still being used (there are lots of better ones) but your example patch would seem to indicate that the answer is yes.

Speaking of other RNG algorithms, I plugged [lp.shhh~], the white noise generator from Litter Power, into your patch and the results were delightfully non-deterministic.

Floating Point's icon

It's a pity that gen~ evidently suffers from the same problem.

A quick look at pink~ shows it doesn't cycle through the same way, which is good.

I'd would be great if c74 could address this in the next incremental.