Optical Feedback techniques

That is very common with laser diodes because they age over time. So do LEDs but AFAIK it's only significant in most applications if you run them close to redline.

While you can obtain a clean optical feedback from a laser diode via its back facet that is another story for LEDs. So the first question would IMHO be, how do you get a meaningful output reading that is not influenced by external light shining back into the LED? Not sure where your reader will be used but outdoors there's bright sunlight and all that. Indoors it's probably more the bling-bling people wear around thir wrists and fingers while holding the reader, in conjunction with bright light sources.

The regulator loop is pretty trivial compared to that part of the job.

Putting _a_ photodetector in the path of the your LEDs to measure _their_ output???

Possibly -- but that assumes that the light output of the three LEDs goes up and down the same in response both to whatever environmental effect (probably temperature and aging) that you're trying to control against, and that it goes up and down the same in response to the LED current.

Or do you have a way to turn on each LED individually?

If you can, you want to adjust each LED output individually.

I hope that you're controlling your LED output by heavily filtered PWM: turning it on and off to get an average is probably not your best path to joy.

Whether you're adjusting them all at once, or individually, I'd read the PD output with an ADC (assuming you can get enough accuracy), and I'd have an integrating controller in there that bumps the PWM up or down until the target output is met.

Putting a PD in your LED's output path? Do you mean in addition to the PD (or whatever) you are using to measure the color (reflection/transmission)?

The best scheme (if possible) uses the same sensor to measure both the LED output and the measured quantity. It is possible to measure the entire optical path attenuation. That way, you can compensate for variations in the optics, the LED and the detector itself.

This is usually done by performing a calibration run of the system before inserting the sample under test. Or switching the optical path between the sample and no sample paths (or a color reference) if you suspect the system drift to be significant over the measurement time.

This will require some sort of mechanics (mirror on an actuator), so it might be beyond a cheap implementation.

Don't know if my ascii art skills are up to this, but...

___________________ clear window ________ _________ heavy felt paper light barrier

PD >| |< LED |< LED |< LED

Basically, the three LEDs shine throught the window through a light sheild. The photodetector would be along the edge of the LEDs light cone, but in the direct path. (an alternative would be to put the PD at a 45 degree angle to get the primary reflection...) You take one reading with no LEDs on, and then readings with each individually on. The actual sample needs to be directly against the window for best accuracy. The LEDs are on a single power supply (switching) set for

Charlie

Hmm, I guess this one became butchered in transmission.

One way that's sometimes done is this:

Window ----------------------------------------- Paper ------------- --------------- | \

Coolness would be a white light source and a color wheel with R, G, B and clear.

If you're willing to pay for a little solenoid-actuated mirror or white card then you could calibrate the whole path -- but boy, the expense would mount!

Yes! We are trying to get this to the lowest price possible for the best accuracy. Moving things, or even reflectors, or pretty much a no-no! I think that Joerg's idea of a black tube might be the most practical. I also have to worry about light reflecting off this calibration PD onto my main measurement PD...

Charlie

The answer to that bit of the problem is that you modulate your LED drive, and synchronously detect the signal that you detect from it.

It won't stop the outside world from saturating your detector, but at least you will know that your detector has been saturated because it's output will be unreasonably high as well as not being modulated.

-- Bill Sloman, Nijmegen

That still doesn't fix self-pollution. Light reflected from some bling-bling on a finger and reflected back inside will be in phase and fool your detector. That's one reason for the design I've sketched.

Reflectors can be super simple. On one design I've helped debug we used snippets of thin Perspex or something like that. That way it's partly translucent and the cost was mere pennies. We didn't even need a hole that way.

have any links to this type of material? I could theoretically use this as a four way splitter, with the target PD looking at right angles to the calibration PD...

Charlie

but

ts

ir

If bling-bling featured corner cube reflectors, it could be a real worry.

-- Bill Sloman, Nijmegen

I don't remember the source, some plastics outlet. But once when somone accidentally jammed a screwdriver in there and busted this 45-degree reflector I didn't have any. I just ripped a see-through hard plastic insert out of a bonbon carton. Worked exactly like the original and we could have cut a hundred or more from that one carton.

Of course, I had to volunteer and eat all those bonbons because the box and their potential freshness was now compromised :-)

If it does then it'll be the expensive kind, the more important the person feels the more karat :-)

me

ure

hing

DC,

but

use

ts

s I

it

and

You can just use a piece of glass as a reflector for the monitor beam. The main beam passes straight through... Then light coming back in reflects the wrong way.

monitor PD up here - Paper ------------- --------------- | piece of glass -> \

I've taken clear acrylic, heated it in a controlled oven to soften it but not burn it, pushed in LED's while it was soft. afterwards, you can clean it up with find emery and then hit it with a heat gun which will then take any abrasion that you in curd to make it turn to glass looking surface.

I made a few optical bezels doing this and I've notice that this material works very well, giving a glossy glass surface.

I've seen prisms made with that stuff just starting with a block of plastic, cut it, sand it smooth as possible and then hit it with a heat gun.

Jamie

You've just reinvented the 'Teleprompter®'. :)

how about analogue feedback using an op-amp?

me

sure

t a

shing

ADC,

but

use

uts

es I

he

he

it

and

te

I've used microscope cover slips as little beam pick-offs. You get a zillion for a few dollars and they are nice and thin so the reflections from the front and back side pretty much over lap.

George H.