This is a yet another contribution to the already extensive literature that surrounds RGB LEDs and changing colours, but in all my research I have not seen the method I have implemented, so here it is in case someone can use it.

I wanted to build a mood lamp type of device, using a single RGB LED placed in a ping pong ball as a diffuser. I needed to find an algorithm that would allow me to fit the code and data in 2k as I am using an ATTiny 2313 chip, prototyped on an Arduino Uno. Most of the code I found used the random() function, ending up over 3k when compiled, probably due to the libraries, and many of the other methods I uncovered just did not do it for me.

The RGB colour space can be visualised as a cube whose (x, y, z) coordinates range from (0, 0, 0) or black, to (255, 255, 255) or white. More generally the cube is defined in the 3D space (0,0,0) to (1,1,1), scaled by 255. The vertices of this cube define the boundaries of the colour space, and moving along the 3D coordinates from one point to the next will naturally provide smooth colour transitions.

The algorithm I have implemented exploits this and traces the RGB coordinates from one corner to another, 'plotting' the coordinates of the line to the RGB led, which displays different colours. In the code below the path of the transitions is defined as series of vertex numbers for each move, but it can equally be implemened by selecting the next vertex randomly. Defining the path can allow a bias for a particular area of the colour cube (eg, cooler or warmer colours), and the path can be much longer than in the code below with little penalty in memory use.

/*

RGB LED - Automatic Smooth Color Cycling

Marco Colli

April 2012

Uses the properties of the RGB Colour Cube

The RGB colour space can be viewed as a cube of colour. If we assume a cube of dimension 1, then the

coordinates of the vertices for the cubve will range from (0,0,0) to (1,1,1) (all black to all white).

The transitions between each vertex will be a smooth colour flow and we can exploit this by using the

path coordinates as the LED transition effect.

*/

// Output pins for PWM

#define R_PIN 3 // Red LED

#define G_PIN 5 // Green LED

#define B_PIN 6 // Blue LED

// Constants for readability are better than magic numbers

// Used to adjust the limits for the LED, especially if it has a lower ON threshold

#define MIN_RGB_VALUE 10 // no smaller than 0.

#define MAX_RGB_VALUE 255 // no bigger than 255.

// Slowing things down we need ...

#define TRANSITION_DELAY 70 // in milliseconds, between individual light changes

#define WAIT_DELAY 500 // in milliseconds, at the end of each traverse

//

// Total traversal time is ((MAX_RGB_VALUE - MIN_RGB_VALUE) * TRANSITION_DELAY) + WAIT_DELAY

// eg, ((255-0)*70)+500 = 18350ms = 18.35s

// Structure to contain a 3D coordinate

typedef struct

{

byte x, y, z;

} coord;

static coord v; // the current rgb coordinates (colour) being displayed

/*

Vertices of a cube

C+----------+G

/| / |

B+---------+F |

| | | | y

|D+-------|--+H ^ 7 z

|/ | / | /

A+---------+E +--->x

*/

const coord vertex[] =

{

//x y z name

{0, 0, 0}, // A or 0

{0, 1, 0}, // B or 1

{0, 1, 1}, // C or 2

{0, 0, 1}, // D or 3

{1, 0, 0}, // E or 4

{1, 1, 0}, // F or 5

{1, 1, 1}, // G or 6

{1, 0, 1} // H or 7

};

/*

A list of vertex numbers encoded 2 per byte.

Hex digits are used as vertices 0-7 fit nicely (3 bits 000-111) and have the same visual

representation as decimal, so bytes 0x12, 0x34 ... should be interpreted as vertex 1 to

v2 to v3 to v4 (ie, one continuous path B to C to D to E).

*/

const byte path[] =

{

0x01, 0x23, 0x76, 0x54, 0x03, 0x21, 0x56, 0x74, // trace the edges

0x13, 0x64, 0x16, 0x02, 0x75, 0x24, 0x35, 0x17, 0x25, 0x70, // do the diagonals

};

#define MAX_PATH_SIZE (sizeof(path)/sizeof(path[0])) // size of the array

void setup()

{

pinMode(R_PIN, OUTPUT); // sets the pins as output

pinMode(G_PIN, OUTPUT);

pinMode(B_PIN, OUTPUT);

}

void traverse(int dx, int dy, int dz)

// Move along the colour line from where we are to the next vertex of the cube.

// The transition is achieved by applying the 'delta' value to the coordinate.

// By definition all the coordinates will complete the transition at the same

// time as we only have one loop index.

{

if ((dx == 0) && (dy == 0) && (dz == 0)) // no point looping if we are staying in the same spot!

return;

for (int i = 0; i < MAX_RGB_VALUE-MIN_RGB_VALUE; i++, v.x += dx, v.y += dy, v.z += dz)

{

// set the colour in the LED

analogWrite(R_PIN, v.x);

analogWrite(G_PIN, v.y);

analogWrite(B_PIN, v.z);

delay(TRANSITION_DELAY); // wait fot the transition delay

}

delay(WAIT_DELAY); // give it an extra rest at the end of the traverse

}

void loop()

{

int v1, v2=0; // the new vertex and the previous one

// initialise the place we start from as the first vertex in the array

v.x = (vertex[v2].x ? MAX_RGB_VALUE : MIN_RGB_VALUE);

v.y = (vertex[v2].y ? MAX_RGB_VALUE : MIN_RGB_VALUE);

v.z = (vertex[v2].z ? MAX_RGB_VALUE : MIN_RGB_VALUE);

// Now just loop through the path, traversing from one point to the next

for (int i = 0; i < 2*MAX_PATH_SIZE; i++)

{

// !! loop index is double what the path index is as it is a nybble index !!

v1 = v2;

if (i&1) // odd number is the second element and ...

v2 = path[i>>1] & 0xf; // ... the bottom nybble (index /2) or ...

else // ... even number is the first element and ...

v2 = path[i>>1] >> 4; // ... the top nybble

traverse(vertex[v2].x-vertex[v1].x,

vertex[v2].y-vertex[v1].y,

vertex[v2].z-vertex[v1].z);

}

}