Measuring color the goofy way

ones, but

the LEDs.

inferior micro here if you must),

the color name on a LCD display.

in the PIC.

Hi Jan, Well, since I have spent the last half of a year trying to do EXACTLY this, I can tell you it ain't so easy!

First the LED's color 'filter' ain't too accruate. Second, response from device to device varies quite a bit, and this really makes your look up table a unit by unit sort of thing. Even if you try mapping RGB to HSL, and working from there, it ain't easy, not easy at all... ;-)

Charlie

Why?

Just use HCFR,

Cheers

On a sunny day (Fri, 25 Dec 2009 10:52:13 -0500) it happened "Martin Riddle" wrote in :

TCS230, interesting chip!

It is discontinued. But there are other offerings....

Cheers

Digikey has an OPB780Z

The Taos stuff is less expensive.

Cheers

My experience seeing spectra of LEDs is that with few exceptions, the color of the plastic has only negligible effect.

With red GaAlAsP (deeper red, peak wavelength generally 660 nm, dominant wavelength generally 640 nm), InGaAlP LEDs including InGaAlP red (peak wavelength generally 630's, dominantwavelength generally 620's), and InGaAlP green and blue LEDs,

I have found that tinting of the LED's body has only negligible effect because the spectrum is already fairly narrow-band. Green and blue ones can be *very slightly* improved. The dyes used are more broadband than the LED emission spectra in these cases.

LEDs where dye in the body *does* have effect that I found significant:

1. Red GaP (doped with ZnO) LEDs, emission peaking at 690-700 nm but more broadband than usual for LEDs, also having a secondary band in the yellowish green that the body's dye can supress. The longer wavelength emission band also has a nonlinear effect, with efficiency decreasing as current increases past a milliamp or two. Without red filtering, these red LEDs often appear orange with an amount of current that they arerated to handle.
2. Yellow-green GaP LEDs - green dying of the body can shift the dominant wavelength (wavelength of single-wavelength light with same hue, at least roughly) shorter by a few nm, making these appear slightly more green and less yellow. (A variation of GaP green also has less doping of some sort to compromise efficiency to make the color less yellowish.)
3. Non-InGaN blue LEDs, which have broader bandwidth than the usual InGaN ones. With these, I have found tinting of the body can make the color a noticeably somewhat more pure shade of blue.

======================================================

Now, as for color sensed by such a scheme: It would sense color of an object as it appears when illuminated by an RGB LED light, not as illuminated by usual white lights or sunlight, daylight, incandescent, xenon, etc. The spectrum of an RGB LED light is low on yellow, blue-green and violet, and extra-rich in red, green and blue. The main effect of this is to make reds appear brighter than "normal", same with greens but to a lesser extent, and oranges, browns and skin tones become more reddish than "normal".

- Don Klipstein ( snipped-for-privacy@misty.com)

ones, but

of the LEDs.

inferior micro here if you must),

reports the color name on a LCD display.

in the PIC.

What about using a "color balanced" fluorescent light source, and 3 photo detectors, each having a different filter (red, green, blue)? Theoretically, one could use two.

On a sunny day (Sat, 26 Dec 2009 00:21:12 +0000 (UTC)) it happened snipped-for-privacy@manx.misty.com (Don Klipstein) wrote in :

Don. great info, thank you! All this makes me think if there are other solutions, somebody mentioned the project with the TCS230 color sensor chip. But even in the data sheet of that chip (page 5), if you look at the curves, it needs an HOYA CM500 IR blocking filter. Given that, and given that most -non security- webcams already have the IR filter, (The security webcams have no IR blocking filter and use IR diodes for night vision) I wonder if I could use my old Creative Videoblaster II webcam, after all I wrote tons of software for it already in the ancient past: