Connect with us

The color reader... again!

Discussion in 'Electronic Design' started by Charlie E., Jun 28, 2011.

Scroll to continue with content
  1. Charlie E.

    Charlie E. Guest

    Hi All,
    Ok, need ya'lls help for a little brainstorming. I have hit a
    calibration snag, and to be honest, it doesn't make much sense.

    I have built 100 units. Of those units, 88 have calibrated just fine.
    However, the other 12 units won't calibrate, and they all have a
    similiar problem, some worse than others, and I have been beating my
    head against the wall for a couple of weeks now...

    In a nutshell, my calibration process is two steps, in which four sets
    of data are taken. First, I calibrate to a known white sample. With
    this, I set the gain for each of my red, green and blue channels to
    get a reading just below max, then store the gain setting and the
    actual max reading.

    I then place the unit in my 'black hole box', a small cardboard box
    lined with black felt, and take a reading. This determines my 'zero
    point' for each color. I change to a 2X gain channel I have, and read
    each color again. I use this for my dark colors.

    Finally, I place a navy blue sample in front of the unit, and store
    the difference between the navy blue and the 2X dark reading. I use
    this as a scaling factor for my dark colors.

    Now, for the mechanics. I have the three LEDs mounted on the top of
    the board. These are side firing LEDS, and they are at a 45 degree
    angle to my window with the red and green LEDs on opposite sides of a
    triangular cut in the board about 1cm from the window, and the blue
    LED about 25 mm farther back. On the bottom of the board I have an
    Ambient Light Sensor (phototransistor based) bent 90 degrees and aimed
    straight at the center of the window. The idea was that the board
    would screen the PT from the LEDs, and that with the 45 degree angle,
    most of the reflection from the window would be off to the side. The
    only signal going to the PT would be the reflectrion from my sample on
    the other side of the window.

    Early in my troubleshooting, I found that I had to color the PCB with
    its green solder mask with a black sharpie to reduce transmission
    through the board, and out the edge of the board. This had
    substantially reduced my dark values, but is one more hand step in the
    assembly process...

    And now, on to the problem. On all the calibrated units, you get a
    pretty set pattern on these dark and navy values, red slightly higher
    than green, which is higher than blue. On the bad units, however, the
    red is MUCH higher than the green, usually in both the dark readings
    and the navy readings.

    My initial thought was that I had mis-aimed the PT so that it was
    seeing more of the red return than the green. Changing the angle of
    the PT does not seem to significantly change the readings. I had
    thought that maybe I had missed a spot on my blacking out the PCB, but
    even recoating that area has no real effect. On one unit, I even put
    in a black construction paper 'sheild' along the bottom of the PCB.
    This did help in further reducing the red dark values, but the high
    read values on the navy sample still remain.

    So, any ideas? Could I have just gotten 12% of the LEDs with much
    higher output? Could my PT be damaged by too high a soldering temp,
    and thereby removed the red filters for the ambient light
    compensation? Right and left field ideas????

    Thanks,
    Charlie
     
  2. Joerg

    Joerg Guest

    Charlie E. wrote:

    [...]

    Now don't hit me, but did you swap the red LED on a bad unit with one on
    a good unit and see if the error "follows along"?


    That would also easily be tested by swapping.

    Take a _very_ close look at a good board and one of the more extreme bad
    ones. It there any component that is crooked, a shiny surface that
    presents itself, or any other stuff that could cause a light path?

    This is ugly but can help: If you pour little globs of goo (yoghurt?)
    onto the board, spot by spot, that should eventually interrupt an
    unwanted light path and cause a drop in reading.
     
  3. Martin Brown

    Martin Brown Guest

    Are all the red LEDs from the same batch ?

    What happens if you measure a bad calibrating ones illumination output
    with a know good unit? Or swap LEDs/PT from a baddie to a goodie?

    Does ambient illumination play any part?

    Regards,
    Martin Brown
     
  4. Charlie E.

    Charlie E. Guest

    Hi Jeorg,
    No, I haven't tried swapping the LEDs out. They were all attached by
    the assembler using automated equipment, so I HOPE they are correct.
    Martin may be right, though, and I may have two different batches of
    LEDs on the boards.

    Those things are tiny, and can be very 'interesting' to solder and
    unsolder. but you are right, I will probalby have to do it. I have
    already tried changing out the PT, which is a major PITA as they are
    through hole parts, and I don't have a good tool for cleaning out the
    holes...

    Charlie
     
  5. Charlie E.

    Charlie E. Guest

    One interesting thought on the multiple batches of LEDs, is that I
    would assume that they were all purchased on a reel, but the problems
    were pretty evenly distributed among the boards as I finished them. If
    they were from a different batch, I would have expected more
    clustering...

    Charlie
     
  6. Joerg

    Joerg Guest

    Swap anyhow, no matter how hard it is. Before you spend more frustrating
    days banging your head against the wall. Then if the error doesn't
    follow along it's probably down to the goo. Yuck. But needs to be done.
    I hope it's not a lamination issue where the light tunnels along inside
    the board.

    As for solder removal get one of those hand-operated pumps. Mine is an
    aluminum version with good seals and it can almost suck a fish straight
    out of the water.
     
  7. John S

    John S Guest

    Order the PCB with black solder mask. See if that helps.

    John
     
  8. Joerg

    Joerg Guest

    I make so few mistakes that I am still on my first braid dispenser since
    highschool ... <gloat .. boast .. blush>

    Truth is, I barely use it and if I need to then it's not right where I
    am sitting. Then us RF dudes use the braid of microphone cable or in a
    pinch coax. Instead of getting up and picking up the dispenser. That's
    how we gain wait.
     
  9. Joerg

    Joerg Guest

    Whoops, sorry. Read something in Spanish, then German, then shifted the
    wrong gears.

    How'd you lose 15#? Fasting?

    I can exercise all day long and ... nada :-(
     
  10. Try measuring with an independent sensor first, then hack away. It's
    the only way to get numbers. WRITE the numbers DOWN (sorry to yell,
    but you need actual numbers to diagnose the problem).

    BTW, did you know you can easily get PCBs made with matt black solder
    mask?



    Best regards,
    Spehro Pefhany
     
  11. Joerg

    Joerg Guest

    You can sometimes also get black circuit boards, meaning the resin in
    the FR-4 is pigmented black.
     
  12. It looks pretty good with gold flash on the boards too. Matt black is
    even better for this application.



    Best regards,
    Spehro Pefhany
     
  13. tm

    tm Guest


    Go to a radio shack and get some solder wick. Or a solder sucker.


    tm
     
  14. Joerg

    Joerg Guest

    Whew, that must have been a scare, especially since you lost Duane to
    colon cancer.

    On account of los doctores or just as a new resolution?
     
  15. Not that you were all that hefty before..



    Best regards,
    Spehro Pefhany
     
  16. Charlie E.

    Charlie E. Guest

    I have a lot of solder wick, which I use a lot of, and I bought one of
    those little rubber bulb things, but by the time you get the bulb to
    the hole, it has already solidified... :-(

    Charlie
     
  17. Charlie E.

    Charlie E. Guest

    Ok, I replace the worst unit with an LED out of my stock from when I
    was prototyping, and the unit calibrates normally. So, it looks like
    I just have a few that are super efficient... :-(

    Of course, I have 12 bad boards, and seven extra LEDs!

    Charlie
     
  18. Cost of 1 LED <<< cost of 1 populated board, right?

    And maybe next time add more adjustment range or a jumpered resistor
    or something like that to get 100% yield without rework.



    Best regards,
    Spehro Pefhany
     
  19. Charlie E.

    Charlie E. Guest

    Yes, hindsight is always great on this. What was funny was that in my
    preliminary prototypes, I had made a mistake and specifiec 32mil thick
    boards instead of the standard 64mils. When I made those up, they
    worked fine. When I went to production, I 'corrected' this to 64mils,
    and found that the phototransistor I was going to use wouldn't fit! Of
    course, I had some smaller ones that I had tried, and those did fit.
    When I get more boards, I will make sure that there is copper pour
    around all the LEDs (2 layer board) to help reduce bleed through. They
    are very near the edge of the board, so I kept the pour back a bit
    from the edge...

    I will also go with the matte black solder mask!

    Charlie
     
  20. Charlie E.

    Charlie E. Guest

    I don't have any good photometric gear, but I am starting to wonder if
    maybe these LEDs might be a little different in wavelength, so they
    get measured differently by the ALS...

    Charlie
     
Ask a Question
Want to reply to this thread or ask your own question?
You'll need to choose a username for the site, which only take a couple of moments (here). After that, you can post your question and our members will help you out.
Electronics Point Logo
Continue to site
Quote of the day

-