Playing with the new new noise functions on a due - full 16 bit noise, 3 octaves for each of hue and brightness, easing on every pixel, and spinning the grid of noise around the center before mapping to LEDs at 80fps (haven’t been able to do arm specific optimizations of the noise code yet, but soon!)
@Daniel_Garcia I was wondering what the benefit is of 16bit noise currently? Because the HSV functions and the ledchips use 8bit values? Or did I miss something?
Each point requires a whole bunch of multiplications, and each one degrades the signal quality a tiny bit. The compound effect is of a loss of some precision. Doing all the math in 16 bits helps preserve the quality of the output. You do all the multiplications in 16 bits, and then at the end you just use the high eight bits, and most of the lossiness is in the low bits, so the output is cleaner and smoother.
If you do just 8 bit math all the way through, there’s more accumulated noise in the result, proportionally speaking.
In some cases, you want the extra smoothness, in some you want the extra speed, so we’re providing both.
Thanks Mark for the clear explanation. That makes sense indeed.