0
\$\begingroup\$

I am trying to make a colorimeter using TCS3200 sensor and white LED light source.

After weeks of research and understanding on how different white light sources work and what are the algorithms involved in calibrating the sensor in CIE1931 space.

Now I am facing problem in getting a reference for getting a calibration curve.

For example, if I want to map R values, then I also need some reference for which I know R values.

I initially planned to use RAL cards to calibrate my sensor, but the RGB values are not shared on their official website and other websites provide sRGB values and not the actual RGB values.

I even thought of printing varying shade of Red on paper, but I believe there will be a lot of unknown factors in this (printer calibration, RGB space used by the software, etc).

So is there any standard source of known colour against which i can calibrate my sensor?

Edit: My application is more to sense the change in shade rather than getting the actual color of the subject and that's the reason I was thinking if I can get away with calibrating it against some standard color swatch of known RGB values.

Improve this question
\$\endgroup\$
  • 4
    \$\begingroup\$ I'm voting to close this question as off-topic because it isn't an EE question. \$\endgroup\$ – Andy aka Jun 19 '17 at 9:16
  • \$\begingroup\$ I understand your point, but can you point me to a right platform where i can ask the same? @Andyaka \$\endgroup\$ – Mayank Jun 19 '17 at 9:27
  • 1
    \$\begingroup\$ There is no translation between subtractive (RAL, NCS, Pantone) to additive (RGB). For a given light source, light level and gamma curve, you can make an approximation. It's like comparing car engine power to top speed for the entire fleet of cars on our roads. They are related, but there isn't a simple formula or linear relationship. \$\endgroup\$ – winny Jun 19 '17 at 10:40
  • 1
    \$\begingroup\$ This is near an area where I specialize (pyrometry [optics and physics of] and human color perception [far more complex subject, but which requires similar physics and optics as grounding.]) I have two primary questions right now: (1) If you build two of these instruments, do you expect reproducability of results between two different devices? (2) Do you seriously imagine that a red LED can make a "standard candle" source? And one comment: You are planning on using RAL color cards, designed for human color perception problems, for a non-human-related problem (glucose content.) \$\endgroup\$ – jonk Jun 19 '17 at 15:04
  • 1
    \$\begingroup\$ @Mayank My email is at the bottom of this page, for example: infinitefactors.org/jonk/patch.html \$\endgroup\$ – jonk Jun 21 '17 at 7:32
2
\$\begingroup\$

You need a light source with a spectrum of known power vs. color (i.e. frequency or wavelength) distribution.

(As much as I understand your problem you don't need absolute intensities (e.g. Watt/Steradian) but just relative intensities with respect to color).

There come two relatively simple sources to mind:

  • black body radiation (e.g. of a incandescent light bulb) whose spectral distribution is directly known by a physical law, Planck's law (you need to know the temperature of the filament)
  • sunlight (after passing through the atmoshphere till e.g. sea level) whose spectral distribution is empirically known (well measured) and you have a high chance of finding published spectral distribution data.
Improve this answer
\$\endgroup\$
  • \$\begingroup\$ That sounds too much of work the project I am looking for. I am looking to measure the change in color intensity hence I believe I might be able to get away without the complexity mentioned above. \$\endgroup\$ – Mayank Jun 19 '17 at 9:32
  • \$\begingroup\$ @Mayank: you should make clear what you actually want to achieve. What do you mean by "color intensisty"? Do you mean saturation ("colorfulness"). Please make sure (or at least put some effort in) using the correct terms. \$\endgroup\$ – Curd Jun 19 '17 at 10:13
  • \$\begingroup\$ Apologies for the confusion, even I am struggling with a lot of names and trying to make sense of it. And yes I mean saturation. \$\endgroup\$ – Mayank Jun 19 '17 at 10:56
1
\$\begingroup\$

long story short: for calibration, you're right, you'll need known quantities. In case of colorimeters, that is something of known color illuminated with light of a known color temperature (and usually, also a known intensity). That's a bit of an expensive test setup you're aiming for, if you need to build it yourself (these lamps don't come cheap).

Why not simply buy a sufficiently calibrated colorimeter, or a colorimeter calibration toolkit, and calibrate your device against the readings / known values of the commercial one? Seems both the cheapest and easiest approach.

Improve this answer
\$\endgroup\$
  • \$\begingroup\$ Actually, my application is more to sense the change in shade rather than getting the actual color of the subject and that's the reason I was thinking if I can get away with calibrating it against some standard color swatch of known RGB values. purchasing colorimeter calibration toolkit seems to be an ideal option given the cost and complexity, though if there is any alternative then please guide with the same. \$\endgroup\$ – Mayank Jun 19 '17 at 9:26
1
\$\begingroup\$

If you're only interested in matching colors you don't need to calibrate. A stable colorimeter will give find differences in color which might, in themselves, be inaccurate, but will show color differences. Your reference should be the sample you're trying to match. The only problem is that you can't match using two different colorimeters.

Improve this answer
\$\endgroup\$

Not the answer you're looking for? Browse other questions tagged or ask your own question.