Question about noise function return ranges: I am extensively using noise8() in my code,

Archive FastLED Archive
1

Question about noise function return ranges:

I am extensively using noise8() in my code, and I found that it seems to return values heavily biased toward the middle of the range (120). I made a loop that iterated 100k times and tested the min/max of noise8() calls using random x and y and z values. I found that the value never dipped below ~60 (this is from memory so could be off a bit), and didn’t generally go all the way up to 255.

Is this by design? In my experience with simplex/perlin noise, it returns values in the full range. I’ve “solved” the problem by rescaling the return value, but it bothers me and I’d rather do it “right”. Thanks!

2

See how @Mark_Kriegsman did it here: https://github.com/FastLED/FastLED/blob/master/examples/NoisePlusPalette/NoisePlusPalette.ino#L98-L100

3

Ah, thanks very much, @Jon_Burroughs !

4

Yeah, it’s a trade off, balance vs. maximizing range - doing all the math in 8/16 bit limits the ranges that I’ve got (and I tried very hard to keep the noise functions from drifting up into 16/32 bit math). Ahh, the joy of fixed point math :slight_smile:

5

That makes perfect sense. The 8-bit math noise functions are a godsend, and they are so fast I can do >2100 calculations per frame and still get >50fps. Thank you so much! My attempts at optimizing simplex noise functions were mediocre and frustrating.