For colours, there's a BIG difference in reproduction on screen (RGB) and in print (CMYK). There are colours that can be represented in one colour space (bright oranges and greens in RGB) that simply don't exist in the other colour space.
But back to OP's problem.
My approach would be:
- find an RGB sensor. Find the spectral response in the data sheet of the three colours.
- find three LEDs, one red, one green, one blue, with a spectrum that matches the sensitivity of the sensor. Bonus for finding them in a single package.
- read the sensor, use the values to drive the brightness of the LEDs.
This should (in theory, at least) give a quite close representation of the sky colour. Some correction factor on the brightnesses may be needed (LEDs may have a different response than the sensor).
vffgaston:
... can you reproduce any colour by mixing yellow, red and blue?
Regards.
Kind of. Here is an 8 part article I wrote back in 2003 about Goethe's theory, colorblindness, and the physiological perception of color (please excuse the wayback links, but that's all I got):
1 - White is Green
2 - Receptors
3 - Color Spaces
4 - Brightness Matching
5 - Color Mixing
6 - Opponent Colors
7 - History
8 - Goethe's Prism
Not all color sensors are born equal.
I spent 10 times more time in coding one than planned, after days of struggling to code another one, which refused to work for misterious reasons. The second one did work, but proved to be useless in practical terms, since color recognition of a plain surface few centimeters in front of the sensor did not make any sense for me.
I made a colour sensor of a simple LDR and a RGB led. It could reproduce a colour using another RGB led. It could pretty well distinguish between different colours "rounding" them to some 6 colours.
@Johan_Ha
Could you please offer more details?
Have an RGB led flash red, green and blue light on an object. Have the LDR read how much light is reflected for each colour. Calibrate.
Johan_Ha:
Have an RGB led flash red, green and blue light on an object. Have the LDR read how much light is reflected for each colour. Calibrate.
Neat.
But hard to flash an LED off the sky.
...R
Nice trick. Won't work for measuring the sky, though.
For representing all colours: if you have a set of three LEDs the spectrum of which perfectly match the sensitivity of the colour receptors in the human eye, then you should be able to recreate all colours humans can distinguish.
First problem: men and women have quite different colour perception. Women can distinguish many more shades in the green/blue/purple parts of the spectrum. That's even before individual differences (which is more than just colour blindness).
wvmarle:
Nice trick. Won't work for measuring the sky, though.
For representing all colours: if you have a set of three LEDs the spectrum of which perfectly match the sensitivity of the colour receptors in the human eye, then you should be able to recreate all colours humans can distinguish.
First problem: men and women have quite different colour perception. Women can distinguish many more shades in the green/blue/purple parts of the spectrum. That's even before individual differences (which is more than just colour blindness).
You don't need to match the LEDs to the eye, matrix math will take care of that. Besides the color response of the S, M and L cones (in trichromatics) overlaps, and are composed of quite complicated curves. Using the CIE diagram (for example,) you can reproduce any color inside of a triangle defined by a set of primary colors as the vertices, with any other set of primaries, given the color being reproduced lies within both triangles.
There is an identified sex-linked gene that gives some percentage of women 4 sets of cones, much as another gene gives some men but two cones (the trit- and deuteranopes.) Now we add the blended retinas, we have a range of men with 2 to 3 cones, and women with 3 to 4 cones, so yes! Many women are better at color than most men. There are outliers, of course.
[edited for grammar and factual clarification]
ChrisTenone:
You don't need to match the LEDs to the eye, matrix math will take care of that. Besides the color response of the S, M and L cones (in trichromatics) overlaps, and are composed of quite complicated curves. Using the CIE diagram, you can reproduce any color inside of a triangle defined by a set of primary colors as the vertices, with any other set of primaries given the color to reproduce lies within both triangles.
It is a known there is a sex-linked gene that gives some percentage of women 4 sets of cones. Hense the belief that women are better at color than men. On average, it's true!
Please, allow me to insist: who's going to accept the result (of the sky remembrance, I mean)?
In reply #8 I suggested a camera and inserting a colour sensor in the place of the film. That would be a simple one pixel digital camera.
Another way would be using a hand torch. Pick one which casts a clear parallel beam. The larger reflector, the better. Replace the light bulb with a colour sensor, but shade direct light from the sensor, allowing only light from the reflector to reach the sensor. With that you can point at objects or at the sky and only light from some spot will reach the sensor. Perhaps adjusting the sensor position will widen the sector (or cone) from which it picks up light.
vffgaston:
Please, allow me to insist: who's going to accept the result (of the sky remembrance, I mean)?
There is a large gallery room, evenly lit with invisible directionless light. The huge room is lined with a couple hundred, 8 foot high, 1 foot wide frames. They go all the way around the room. There are no breaks in the wall. Each frame is illuminated with a slightly different color, defined by a colorimetric sample of the sky. The viewers walk around the room - or the room walks around them, you can't tell - endlessly. There is never enough differentiation to de-mark a beginning or an end to the stream.
(in)finte.
The reflector idea won't work well as colour sensors are pretty directional by themselves. The reflector would either need an omnidirectional sensor, or a second inverted reflector that goes in place of the light bulb (the focal point of the reflector), reflecting the light as a parallel beam on to where the base of the bulb was. That then can be picked up by a colour sensor. A short narrow tube around the sensor, and the sensor moved back and away from the reflector a bit should take care of most ambient light, at least to the extent that it's not a significant influence any more.
wvmarle:
The reflector idea won't work well as colour sensors are pretty directional by themselves. The reflector would either need an omnidirectional sensor, or a second inverted reflector that goes in place of the light bulb (the focal point of the reflector), reflecting the light as a parallel beam on to where the base of the bulb was. That then can be picked up by a colour sensor. A short narrow tube around the sensor, and the sensor moved back and away from the reflector a bit should take care of most ambient light, at least to the extent that it's not a significant influence any more.
Plus you would want to remove any interferences like clouds, trees or parts of whatever is holding the sensor. a 15 degree circle is popular among various camera's light meters (which have accounted for color since the late 70s.) Sky color is usually measured in kelvins, reducing less actinic stray light to single digit percentages.