More questions on color reader...

On the turn on/turn off times, I use a nop timer with 64K counts on a

20MHz clock, so 3 to 10 ms. I start the LED, wait, start the power supply, wait, and then measure four times and divide by two. Turn off the power, wait, and start over...

Charlie

[...]

[...]

That's done very well. This pretty much leaves age-drift of the LEDs themselves, maybe they are just pushed too hard. Possibly only one of them is. A way to test that would be to equip the unit that has crept up past the 240 range with a fresh battery and see if it comes back down.

Can you ease off a bit to, say, half the current and see if the longterm stability improves? It would be sad to throw in the towel at the last mile. Any chance to borrow a digital scope?

--
Regards, Joerg

http://www.analogconsultants.com/

"gmail" domain blocked because of excessive spam.
Use another domain or send PM.

Please look for my PM.

1rst, like Joerg, I'm not happy that IR is being excluded from the detector. Many optically opaque, even black materials are like glass to IR. Use aluminum foil, then you'll know for sure.

The phototransistor filter is, well, pathetic, not nearly good enough. Consider:

The sun produces 1KW/m^2. That's 1mW/mm^2.

Just wildly ballparking with very generous assumptions, your LEDs, at

20mA, produce possibly 30% x 20mA x 3v =3D 18mW over a 10mm spot.

That's 18mW / 76mm^2 =3D 230uW/mm^2 _radiated_; the reflected signal you receive will be an order of magnitude weaker.

So the fact that your sensor's filter is 75% down by 900nm is basically no protection from sunlight at all--sunlight could easily produce 10x more response than your signal and swamp your sensor, if you let it in.

2nd, simplify the problem. For example, o indirectly couple one LED to the phototransistor, attenuated, totally excluding external light, and see if the readings drift. Heat and cool things too. o Measure the actual voltage at the phototransistor's emitter when you do this--see how close to the margins it's getting pushed. o Put the whole thing in a dark room with a controlled light source, if need be, then try turning on an interfering light source, like a fluorescent light, then an incandescent. o Check the raw readings and corrected--see whether the channels respond the same ways. (For example, your phototransistor's response to blue light is down 70%--nearly as far as for 900nm IR--so the blue channel might be more susceptible to the effects of stray light.)

Poke around, think, explore. You'll find it, learn something in the doing, and your product will sing.

You *can* fix it Charlie!

-- Cheers, James Arthur

Yup, got to try foil and stuff. Plastic is (mostly) not very good at muffling IR. The fact that Charlie is seeing drift that "sticks", IOW that doesn't come back after cooling off, tells me that the budget above may have to get worse. He may have to back off on the LED excitation current. Maybe to 10mA or so.

[...]

I'll second that. With enough determination I am sure he can solve this and leave others who possibly don't know about the IR effects in the dust.

--
Regards, Joerg

http://www.analogconsultants.com/

"gmail" domain blocked because of excessive spam.
Use another domain or send PM.

y , 5% u

RFI's a possible too. I'm skeptical of LED aging at 20mA, but that's easily tested.

-- Cheers, James Arthur

True, RFI can do this but it wouldn't "stick".

--
Regards, Joerg

http://www.analogconsultants.com/

"gmail" domain blocked because of excessive spam.
Use another domain or send PM.