RGB Color-Mixing?

[P. Short]

Has anyone done any semi-serious work on color calibration of any particular RGB LED? While it seems to me that almost every RGB LED out there has different luminous intensity ratings for the various component LEDs, I would still be interested in knowing about any testing or work in this area.

I'm starting more or less from scratch in learning about this, and am looking for whatever information I can find. I have a design using PIC10F222 chips for PWM control of RGB LEDs with a more-or-less exponential response curve, and am looking to select some LEDs and then calibrate them as best I can.

[aussiephil]

Phil

semi serious - yes, any useful sources of info - not really. Most of the knowledge comes from past work in digital imaging and colour calibration of the entire image chain.
I have parts to build a light intensity measuring and hardware curve generation based on the standard 255 levels in DMX, just need time......
I have commercial colour calibration and measuring gear used for CRT/LCD/Projector colour calibration but never thought it required to calibrate any of the RGB stuff i'm working on, though the Floodlight design in my head would benefit from it.

In terms of generating White from RGB the key thing is the final colour temperature you desire.
what is considered "normal" white for an incandescant bulb is actually quite a warm (reddish) colour with most RGB LEDS actually mixing a much cooler (blueish) white, this is fairly easy to tweak at max intensity but harder to do over a full dimming curve.

The other interesting trap is just how your eyes perceive brightness, measuring the actual intensity of light output of the 5050 Leds i'm using the LUX reading from 10->225 levels in vixen was near linear (straight line) yet my eyes told me i had full brightness at around level 70. This was interesting and informative.

Maybe i can answer questions you may have.

Cheers
Aussiephil

[djulien]

Has anyone done any semi-serious work on color calibration of any particular RGB LED?

By "calibration", are you refering to how well the colors mix when viewed by the human eye, by a digital camera, or compared to some other source (such as a computer screen or published sets of wavelengths)?

don

[budude]

The problem I've seen is in the inconsistency of most RGB LEDs. After soldering up all my LED strips using Superflux LEDs I noticed that even with the same power source and using 1% drop resistors that the color varied considerably and with LEDs presumably from the same lot. I think a lot of this comes from the diffusion (or lack thereof) of the plastic encasing the LED chips. Also - with typical 3-chip LEDs there are pobably minor differences in the planes of the chips and the near infinite viewing angles (like viewing it from the "red" side versus the "blue" side).

[P. Short]

In terms of generating White from RGB the key thing is the final colour temperature you desire.
what is considered "normal" white for an incandescant bulb is actually quite a warm (reddish) colour with most RGB LEDS actually mixing a much cooler (blueish) white, this is fairly easy to tweak at max intensity but harder to do over a full dimming curve.

Maintaining some color consistency over a full dimming curve (or at least a wide range) is something that I'd like to achieve.


The other interesting trap is just how your eyes perceive brightness, measuring the actual intensity of light output of the 5050 Leds i'm using the LUX reading from 10->225 levels in vixen was near linear (straight line) yet my eyes told me i had full brightness at around level 70. This was interesting and informative.

This is why I'm playing around with a quasi-exponential curve. I also noticed that there was not a lot of response at the upper end with a linear curve.


By "calibration", are you refering to how well the colors mix when viewed by the human eye, by a digital camera, or compared to some other source (such as a computer screen or published sets of wavelengths)?


For the moment, by the eye.

So far I've just been using some LFL550 LEDs that I had around, although I only have a few and they're not longer available. The issue with these is that the various colors are still visibly separate, even though they are milk-white diffused LEDs. I'm planning on ordering some SFT722N LEDs (5050, I think) from mouser, in the hope that they will be more consistent and I can come up with some scheme so that the different color LEDs are not visibly separate.

[aussiephil]

The problem I've seen is in the inconsistency of most RGB LEDs. After soldering up all my LED strips using Superflux LEDs I noticed that even with the same power source and using 1% drop resistors that the color varied considerably and with LEDs presumably from the same lot. I think a lot of this comes from the diffusion (or lack thereof) of the plastic encasing the LED chips. Also - with typical 3-chip LEDs there are pobably minor differences in the planes of the chips and the near infinite viewing angles (like viewing it from the "red" side versus the "blue" side).

All LEDs should (repeat should) be driven from constant current sources, the differences you will seeing is due to different Vf (forward voltage) drops led to led that unless you match each led to the correct value resistor is unavoidable when setting current with resistor values.
Water clear casings do not help the matter as you can see the individual chips and we are drawn to seperate the points.

Phil - something that may be useful
http://www.sciencedirect.com/science?_ob=ArticleURL&_udi=B6V01-4VTKKSN-1&_user=10&_coverDate=07%2F31%2F2009&_rdoc=1&_fmt=high&_orig=search&_sort=d&_docanchor=&view=c&_searchStrId=1196535710&_rerunOrigin=google&_acct=C000050221&_version=1&_urlVersion=0&_userid=10&md5=986b6f7699b63aee0941b79eff19e45c

Cheers

[aussiephil]

Phil

Here's what you need
http://www.pce-group-europe.com/english/index.php/cat/c1015_Colour%20Meters%20PCE-RGB%202.html/XTCsid/53c86740e00ea983a59a1265e0e08685

the SFT722N LEDs is not a 5050 package as it is 6mmx5mm the 5050's are 5mm square.

Cheers

[P. Short]

How well do the color testers do with PWM modulated light? Do they do time-averaging in a similar manner with human eye response?

[mrpackethead]

Somethign else to chuck into the mix.. The color is skewed by the temperture of the led dye. As a general rule, if you go over 70C you're goign to not only dramatically reduce the life time of the led, you'll also get color shift.

Not such an issue with small leds, but if you crank up the power, things start getting all wobbly.

[Entropy]

This is why I'm playing around with a quasi-exponential curve. I also noticed that there was not a lot of response at the upper end with a linear curve.

This is why the microcontroller firmware in my I2C project does a gamma curve expansion lookup, so it has a brightness response similar to how most PC monitors and TVs work. (sRGB color space is gamma-compressed.) - the 256 input values get mapped to 12-bit sigma-delta modulator inputs.



For the moment, by the eye.

So far I've just been using some LFL550 LEDs that I had around, although I only have a few and they're not longer available. The issue with these is that the various colors are still visibly separate, even though they are milk-white diffused LEDs. I'm planning on ordering some SFT722N LEDs (5050, I think) from mouser, in the hope that they will be more consistent and I can come up with some scheme so that the different color LEDs are not visibly separate.
I have been using 10mm diffused LEDs from Evil Mad Scientist Labs. They mix color pretty well, far better than the single 5mm unit from an electronics store in NJ I initially prototyped with.

One thing I have observed is that I get a reasonable "cool white" when all resistors are of equal value with my current LEDs, so moving to values that provide full drive to each channel (as opposed to significantly more to red than G/B due to red having a lower Vf) may not prove to be the best option. I'm worried about unit-to-unit variance when I scale things up though.