Expected range for stb_perlin_fbm_noise3

[jdoege]

What is the expected range for stb_perlin_fbm_noise3? I thought perlin noise was supposed to lie between -1 and 1, but I've seen values from -1.2342147 to 1.1876396. To be able to scale the output, it is useful to have a notion of the bounds.

[nothings]

https://digitalfreepen.com/2017/06/20/range-perlin-noise.html

[nothings]

That's just for the base noise function. One point of perlin noise is that it is Gaussian-like, which means there ARE supposed to be rare extremal values, it doesn't go sharply to 0 at the ends. Of course it's a simple finite implementation so it's not actually infinite in extent, but it's AFAIK intentionally not well-defined because of this property. Once you introduce fbm to it, it's summing multiple noises, so it depends on how many octaves and their relative scaling, and the maximum will occur when all of them "constructively interfere", which will be incredibly rare.

[jdoege]

Thank you. That was helpful. Running against stb_perlin_noise3, the bounds stay within -1 and 1. I suspect the outside bounds for stb_perlin_fbm_noise3 so long as lacunarity is 2 and gain is .5 should be -2 to 2 regardless of the number of octaves (1+1/2+1/4+1/8... trends to 2). Change lacunarity and gain and it becomes more complicated.

[jdoege]

And, looking a bit further, I am still a bit off since that link suggests the range for 3d perlin noise is -.866 - .866. That, however, is not what I see running stb_perlin_noise3. After generating a billion values, I get a min of -.898 and max of .689 the lesser of which is a bit out of bounds.