kriegsman

EXPLANTION: This table shows the total light output for FastLED v2.1's "hsv2rgb_rainbow" function 
for a variety of hue, saturation, and value inputs.  
Each cell shows the sum of the Red, Green, and Blue values returned from the HSV-to-RGB conversion, 
which we could call the 'total numeric light output' for that RGB color.

FINDINGS: For any given saturation and value, the 'total numeric light output' is basically constant 
across all hues +/-1, with the exception of yellow and the hues near yellow, which have higher 
power output -- by design.

NEXT: A question remains though: does "252,0,0" _appear_ brighter or dimmer than "126,126,0", 

kriegsman

hsv2rgb_spectrum 'total numeric light output' across hues, saturations, and values.
(Same as hsv2rgb_raw, but with hues 0..255, instead of 0..191)
Each cell represents the total of R+G+B for the output of the hsv2rb conversion.

Findings: 'total numeric light output' is constant across all hues, +/-1.
N.B. this was, in fact, one of the target design paramaters of this algorithm.

Remaining open question: does 252,0,0 appear exactly as bright, or brighter, or
dimmer than 126,126,0 or 84,84,84, even though they all have the same 'total
numeric light output'.

kriegsman

hsv2rgb_raw_kasper 'total numeric light output' data.

This data is somewhat similar to the output of traditional hsv2rgb 
algorithms, such as the one described on wikipedia, or Adafruit's 
color "wheel" function.  In this model, blended hues requiring a mix 
of LEDs (e.g, aqua, hue=96, output=510) output almost twice as much 
light as primary hues that use just one LED, (e.g. blue, hue=128, 
output=255).  This can result in the blended hues appearing brighter 
than primary hues.  In the case of Aqua, this algorithm turns the 
Green LED on to 254 AND the Blue LED on to 254: two LEDs on at full power.  

kriegsman

0008-d150:  da fe d1 fe-d2 fe cb fe-c8 fe c2 fe-bf fe ba fe  ........ ........
0008-d160:  b6 fe b3 fe-ac fe aa fe-a4 fe 9f fe-9d fe 97 fe  ........ ........
0008-d170:  92 fe 8f fe-89 fe 85 fe-80 fe 7c fe-77 fe 73 fe  ........ ..|.w.s.
0008-d180:  6d fe 69 fe-65 fe 5e fe-5c fe 55 fe-51 fe 4f fe  m.i.e.^. \.U.Q.O.
0008-d190:  43 fe 46 fe-3b fe 3b fe-33 fe 2e fe-2c fe 22 fe  C.F.;.;. 3...,.".
0008-d1a0:  23 fe 18 fe-19 fe 0e fe-0e fe 06 fe-00 fe 01 fe  #....... ........
0008-d1b0:  f2 fd f8 fd-e8 fd ec fd-e0 fd e0 fd-d7 fd d4 fd  ........ ........
0008-d1c0:  cd fd ca fd-c3 fd be fd-b9 fd b4 fd-ae fd aa fd  ........ ........
0008-d1d0:  a3 fd 9f fd-99 fd 93 fd-90 fd 87 fd-85 fd 7e fd  ........ ......~.
0008-d1e0:  79 fd 73 fd-6e fd 69 fd-62 fd 60 fd-56 fd 54 fd  y.s.n.i. b.`.V.T.

kriegsman

#include <float.h>
#include <stdio.h>
#include <string.h>

// Output from gcc on Intel x64, all default settings:                                                                                
// - float's can be up to 46 characters in %f output,                                                                            
// - double's can be up to 316 characters in %lf output.                                                                         
//                                                                                                                               
//                                                                                                                               
//   printf("%f",  FLT_MAX) = 340282346638528859811704183484516925440.000000, (46 chars)                                         

kriegsman

In order to evaluate the true randomness of "random8", I generated 64K of random bytes using random8, and ran it through "ent" to analyze the entropy:

  Entropy = 8.000000 bits per byte.

  Optimum compression would reduce the size
  of this 65536 byte file by 0 percent.

  Chi square distribution for 65536 samples is 0.00, and randomly
  would exceed this value more than than 99.99 percent of the times.

kriegsman

#include <FastLED.h>

#define LED_PIN     3
#define LED_TYPE    WS2811
#define COLOR_ORDER RGB
#define COLOR_CORRECTION CRGB(255,172,240)

#define BRIGHTNESS  64

#define NUM_LEDS    100

kriegsman

// See also __uqadd8

static inline uint8x4_t uqadd8x4 (uint8x4_t xs, uint8x4_t ys)
{
    uint8x4_t sums;
    asm volatile(
        "uqadd8 %[sums], %[xs], %[ys]" 
        : [sums] "=r" (sums) 
        : [xs] "%r" (xs), [ys] "r" (ys));
    return sums;

kriegsman

#include "FastLED.h"

// Quick and dirty 2-D fireworks simulation
// Originaly designed (and tuned for) an Adafruit 5x8 WS2811 shield
// by Mark Kriegsman, July 2013
// (and not updated since, so it's a little stale, 
//  but it's a good starting point, if rather uncommented.)

#define NUM_LEDS 40
CRGB leds[NUM_LEDS];

kriegsman

(Code added to FastLED3.1 branch... find it there.)