Hi! I experimented with the 1D 8 bit noise function and created this variant

Archive FastLED Archive
,
1

Hi!
I experimented with the 1D 8 bit noise function and created this variant that directly generates noise in the range -128…127. No need to scale it, i.e. no _raw variant. The number of multiplications is the same for the not _raw variant, but the result looks very smooth. What do you think?

// generate noise in the range -128…127 for x in 8.8 fixed point notation
inline int8_t noise8s(uint16_t x) {
// integer part of x
uint8_t i = x >> 8;

// fractional part of x
uint8_t f = uint8_t(x);

// bump for position i and negative bump for next position
int8_t g1 = bump8s(i, f);
int8_t g2 = ~bump8s(i + 1, f);

// interpolate between the bumps
return lerp8s(g1, g2, f);

}

// average for 8 bit signed integers
inline int8_t avg8s(int8_t i, int8_t j) {
return (i + j + 1) >> 1;
}

// lerp for 8 bit signed integers. the interpolator r is interpreted as 0…255/256
inline int8_t lerp8s(int8_t a, int8_t b, uint8_t r) {
if (b > a)
return a + uint8_t(r * uint8_t(b - a) >> 8);
return a - uint8_t(r * uint8_t(a - b) >> 8);
}

// calculate bumps of varying height in the range -128…127
inline int8_t bump8s(uint8_t i, uint8_t x) {
int8_t hash = P(i);
int8_t y = uint8_t(x * x >> 7) - uint8_t(x + x);

int8_t u, v;
if (hash & 8) {
	u = y;
	v = y;
} else {
	if (hash & 4) {
		u = 1;
		v = y;
	} else {
		u = y;
		v = 1;
	}
}

if (hash & 1)
	u = ~u;
if (hash & 2)
	v = ~v;

return avg8s(u, v);

}

5656183d704f13215a903826a22899bd.png
5656183d704f13215a903826a22899bd.png1280×200 25 KB

2

Looks good @Jochen_Wilhelmy_Digi ​. Thank you for posting.

3

The following line causes the code to not compile:

int8_t hash = P(i);

In the meantime, it looks like the input is an unsigned 16 bit value, which I assume you just increment or something, while the output is that 8 bit signed value.

I personally prefer a uint8_t returned value, but then my programming is amateur hour, so I’m probably missing something important here.

4

The P(i) is the permutation table from fastled’s noise (http://fastled.io/docs/3.1/noise_8cpp_source.html). Of course the output range can be shifted to 0…255 by adding 128