Shop OBEX P1 Docs P2 Docs Learn Events
TCS3200 emissive response — Parallax Forums

TCS3200 emissive response

bkirkbkirk Posts: 37
edited 2013-05-31 07:05 in Accessories
In the documentation for the TCS3200 it has a graph showing the relative response for emissive light. I'm trying to figure out how to compensate for the differences between the photodiodes when using a constant time across all 4 of the photodiodes. For example, if I'm using 100 as the time for all 4, how do I determine, based on the chart, how to adjust the color photodiodes to balance out their relative responsiveness and get an accurate reading of their values in that time of 100? Thanks.

Comments

  • Phil Pilgrim (PhiPi)Phil Pilgrim (PhiPi) Posts: 23,514
    edited 2013-05-22 16:16
    What you need to do is to determine the area under each curve and multiply the response by a number inversely proportional to that area for each of red, green, and blue.

    It's much simpler just to do a white balance if you have a known white subject, though.

    -Phil
  • bkirkbkirk Posts: 37
    edited 2013-05-22 17:30
    Under the curve for each color, or the area between the clear curve and the colors?

    I have a high quality white LED light; can I direct that into the sensor and calibrate the colors? The goal would be to just make them equal, right?
  • Phil Pilgrim (PhiPi)Phil Pilgrim (PhiPi) Posts: 23,514
    edited 2013-05-22 20:54
    bkirk wrote:
    Under the curve for each color, or the area between the clear curve and the colors?
    Under the curve for each color.
    I have a high quality white LED light; can I direct that into the sensor and calibrate the colors?
    Yes, so long as it uses broad-spectrum fluorescence to generate the white light and not just RGB LED elements. An incandescent bulb would be better, though, since it has a smooth, even spectrum.
    The goal would be to just make them equal, right?
    Right.

    -Phil
  • bkirkbkirk Posts: 37
    edited 2013-05-23 06:57
    Cool, thank you!
  • bkirkbkirk Posts: 37
    edited 2013-05-31 07:05
    Here's what I found, and figured I should share (hoping it's correct).

    I figured out the area below each color's curve first. I did so by counting the squares and fractions of squares thereof. I found 18 for red, 15 for green, and 13 for blue. I then took readings under a broad spectrum light bulb. I multiplied the responses by the inverse of the above numbers.

    The results were a little off, as my method of area estimation was, well, estimation. I made some adjustments using the light and found that 17.5 for red, 14.5 for green, and 13 for blue evened out the responses nicely. When I took the inverse I got .057 for red, .069 for green, and .077 for blue. I don't know why I needed to do this (maybe someone can explain why; I imagine it has to do with the chart being fractions of 1, or something), but I then multiplied them all by 10, because otherwise the responses were way too low relative to the clear response.

    The final result was .57 for red, .69 for green, and .77 for blue. This appeared to balance the responses pretty well under the light, made them all make sense relative to the clear response (in total the 3 equaled about 75% of the clear response) and made sense based on the areas under each curve that I calculated.

    Hope this helps anyone needing to calibrate based on emissive light.

    Ben
Sign In or Register to comment.